chore: move scripts and runtime sources under litellm

.agents/skills/litellm-setup/SKILL.md

@@ -19,10 +19,10 @@ Ask everything in one message, with defaults the user can accept by saying "go":
 
 - Providers and models. Default is OpenAI with its newest flagship model, using the `OPENAI_API_KEY` in your environment when `[ -n "$OPENAI_API_KEY" ] && echo set` says it is there. Run that check before asking, and say in the question that the default uses that variable, naming it but never its value, since a key in your environment can belong to another account
 - Whether they want virtual keys, teams, spend tracking, or the Admin UI. Those need Postgres, plain routing does not. Default is no
-- uv (default) or Docker. With Postgres, Docker runs the repo's quickstart, which brings its own database, and uv needs a Postgres server they already have, with a new, empty database for LiteLLM, so someone with their own Postgres server picks uv
-- Folder and port. Defaults are `./litellm-proxy` (`./litellm-gateway` for the quickstart) and 4000
+- uv (default) or Docker. For Docker with Postgres, use the deployment guide in step 3. uv needs a Postgres server they already have, with a new, empty database for LiteLLM
+- Folder and port. Defaults are `./litellm-proxy` and 4000
 
-Then follow the steps for their answers. Plain routing skips steps 3 and 6. The quickstart is steps 2, 3, 8, and 9. uv with their own Postgres skips step 3
+Then follow the steps for their answers. Plain routing skips steps 3 and 6. uv with their own Postgres skips step 3
 
 In the same message, say that someone who picks uv and would rather answer prompts can run `litellm --setup` in a separate terminal window once LiteLLM is installed, not through Claude Code. Never run it yourself, since it needs a real terminal and without one prints "Setup cancelled." and writes nothing. Give them the command as `LITELLM_MODE=PRODUCTION <bin>/litellm --setup`, with `<bin>` replaced by the folder `uv tool dir --bin` prints, to run from the folder they want the config in. It writes `litellm_config.yaml` there with the provider keys and master key in plain text and offers to start the proxy, which then listens on every network interface and stops with Ctrl+C in that window. When the user is done, run this in that folder. It makes the file readable only by them, keeps it out of git (the wizard does neither), and prints only the model names. Never print that file or run `grep`, `sed`, or `cat` on it:
 
@@ -43,29 +43,13 @@ export LITELLM_MASTER_KEY="$("$(uv tool dir)/litellm/bin/python" -I -c 'import y
 uv --version; docker info --format '{{.ServerVersion}}'; docker compose version; curl --version | head -1; openssl version; command -v python3
 ```
 
-On native Windows, ask the user to run Claude Code inside WSL, since every command here is POSIX shell. Every path needs curl, openssl, and python3, so ask the user to install any that is missing. The uv path needs uv, which downloads a supported Python (LiteLLM supports 3.10 to 3.14) when the `--python` range in step 4 rules out the one on the machine. If uv is missing, offer to install it, and only after the user says yes run `curl -LsSf https://astral.sh/uv/install.sh | env UV_NO_MODIFY_PATH=1 sh`, which leaves their shell profile alone (https://docs.astral.sh/uv/getting-started/installation/). It puts uv in `~/.local/bin`, so start every later command that runs `uv` or `litellm` with `export PATH="$HOME/.local/bin:$PATH";`. The Docker path needs a running Docker daemon, and the quickstart also needs Docker Compose v2. If `docker info` says `permission denied`, the user's account cannot use Docker without sudo. Offer the uv path, or ask them to add themselves to the `docker` group the way https://docs.docker.com/engine/install/linux-postinstall/ shows and start a new login session before you check again
+On native Windows, ask the user to run Claude Code inside WSL, since every command here is POSIX shell. Every path needs curl, openssl, and python3, so ask the user to install any that is missing. The uv path needs uv, which downloads a supported Python (LiteLLM supports 3.10 to 3.14) when the `--python` range in step 4 rules out the one on the machine. If uv is missing, offer to install it, and only after the user says yes run `curl -LsSf https://astral.sh/uv/install.sh | env UV_NO_MODIFY_PATH=1 sh`, which leaves their shell profile alone (https://docs.astral.sh/uv/getting-started/installation/). It puts uv in `~/.local/bin`, so start every later command that runs `uv` or `litellm` with `export PATH="$HOME/.local/bin:$PATH";`. The Docker path needs a running Docker daemon, and Docker with Postgres also needs Docker Compose v2. If `docker info` says `permission denied`, the user's account cannot use Docker without sudo. Offer the uv path, or ask them to add themselves to the `docker` group the way https://docs.docker.com/engine/install/linux-postinstall/ shows and start a new login session before you check again
 
 Check the port is free with `curl -s -o /dev/null --max-time 2 http://127.0.0.1:4000; echo $?`, which prints `7` (could not connect) only when nothing listens there. Don't use `lsof` for this, since it misses ports that root or Docker hold. If it is taken, pick another port and use it everywhere below, because a proxy asked for a busy 4000 silently moves to a random port
 
-## 3. Postgres with Docker: run the quickstart
+## 3. Postgres with Docker
 
-The repo's quickstart starts the gateway, Postgres, and the Admin UI in Docker, listening on this machine only, and puts a `litellm-gateway` command in that folder that stops, starts, updates, or removes them. It asks nothing when it has no terminal. It writes the master key, salt key, and database password to its own `.env` and never prints them. Download it, then run it with the folder and port from step 1. It returns once the gateway is ready:
-
-```bash
-curl -fsSL https://raw.githubusercontent.com/BerriAI/litellm/main/scripts/quickstart.sh -o quickstart.sh
-LITELLM_DIR="$PWD/litellm-gateway" LITELLM_PORT=4000 sh quickstart.sh --yes
-cd litellm-gateway
-```
-
-Models live in the database here, so add each one through the API instead of a `config.yaml`, naming it the way step 5 says. Pipe the request body in, so the provider key never shows up in output or on a command line. Unless the user agreed to the key in your environment, as step 5 says, ask them to add the line `OPENAI_API_KEY=<their key>` to this `.env` themselves. Run the loader line first:
-
-```bash
-python3 -c 'import json, os, sys; json.dump({"model_name": "gpt-6.1-sol", "litellm_params": {"model": "openai/gpt-6.1-sol", "api_key": os.environ["OPENAI_API_KEY"]}}, sys.stdout)' \
-  | curl -sS -o /dev/null -w '%{http_code}\n' http://127.0.0.1:4000/model/new \
-      -H "Authorization: Bearer $LITELLM_MASTER_KEY" -H "Content-Type: application/json" -d @-
-```
-
-Run that in one shell command that starts with the loader line, since the master key is in `.env`. `200` means the model was added. Then go to step 8
+Follow [the container deployment guide](../../../litellm/deploy/docker/README.md) instead of the remaining steps in this skill. Deployment tooling is maintained separately from these local uv and standalone Docker instructions.
 
 ## 4. Install
 
@@ -170,7 +154,7 @@ Readiness reports `"db":"connected"` with Postgres and `"db":"Not connected"` wi
 
 ## 8. Verify with a chat completion
 
-Use a `model_name` from `config.yaml` (or step 3), in one shell command that starts with the loader line:
+Use a `model_name` from `config.yaml`, in one shell command that starts with the loader line:
 
 ```bash
 curl -sS http://127.0.0.1:4000/v1/chat/completions \
@@ -186,7 +170,7 @@ Tell the user, without printing any secret:
 
 - The base URL is `http://127.0.0.1:4000`. Any OpenAI-compatible client works with it as the base URL and the master key or a virtual key as the API key
 - The master key is `LITELLM_MASTER_KEY` in the folder's `.env`, or `general_settings.master_key` in `litellm_config.yaml` when they used `litellm --setup`
-- To stop it, `kill "$(cat litellm.pid)"` (uv), `docker stop litellm` (Docker), `<folder>/litellm-gateway stop` (quickstart), or Ctrl+C in its window (`litellm --setup`). To start it again, `docker start litellm`, `<folder>/litellm-gateway start`, on uv the whole step 7 command as you ran it, loader line and PATH export included, wrapped in `( ... )` so their shell keeps its own settings, shown as a code block and never paraphrased, since a new shell has neither, or for `litellm --setup` `LITELLM_MODE=PRODUCTION <bin>/litellm --config litellm_config.yaml --host 127.0.0.1 --port <port>` in a terminal window in that folder, with `<bin>` and `<port>` filled in, since running the wizard again writes a new config with a new master key. `update` and `uninstall` update or remove a quickstart install. Give the quickstart commands with the folder's full path, since a bare `litellm-gateway` is on PATH only when `~/.local/bin` is, and can belong to an earlier install
+- To stop it, `kill "$(cat litellm.pid)"` (uv), `docker stop litellm` (Docker), or Ctrl+C in its window (`litellm --setup`). To start it again, `docker start litellm`, on uv the whole step 7 command as you ran it, loader line and PATH export included, wrapped in `( ... )` so their shell keeps its own settings, shown as a code block and never paraphrased, since a new shell has neither, or for `litellm --setup` `LITELLM_MODE=PRODUCTION <bin>/litellm --config litellm_config.yaml --host 127.0.0.1 --port <port>` in a terminal window in that folder, with `<bin>` and `<port>` filled in, since running the wizard again writes a new config with a new master key.
 - With Postgres, the Admin UI is at `http://127.0.0.1:4000/ui` (user `admin`, password the master key), and `POST /key/generate` with the master key creates virtual keys with their own budgets and model lists. Without Postgres, both need step 3 or 6 first
 - The Admin UI warns that the master-key login is a shared secret, which is fine while only this machine reaches the proxy. Before anyone else can reach it, invite an admin with their own password on the Internal Users page, then set `general_settings.disable_env_credential_login: true` in the proxy config
 - To manage the running proxy from a coding agent (keys, teams, users, models, MCP servers), install `npx skills add BerriAI/litellm-skills -g`
@@ -196,10 +180,10 @@ Tell the user, without printing any secret:
 - `refused to start: no master key is set` means the loader line did not run in the shell command that started `litellm`, or `--env-file .env` is missing from `docker run`
 - `a database URL is set but the prisma CLI is neither on PATH nor importable` means the install lacks `extra_proxy`, so rerun step 4 with it, or, on plain routing, that the command that started `litellm` did not start with the loader line
 - `The Client hasn't been generated yet` or `Unable to find Prisma binaries` means the Prisma generate command in step 6 has not run
-- `Incorrect API key provided` or another provider `AuthenticationError` means a provider key is wrong. Fix it in `.env` and restart, since the proxy reads the environment only at startup. On uv, stop it and rerun step 7. On Docker, `docker rm -f litellm` and rerun step 7, since `docker restart` keeps the environment the container was created with. On the quickstart the key is saved with the model, so delete the model and add it again: `GET /model/info` lists each model's `model_info.id`, and `POST /model/delete` with `{"id": "<id>"}` deletes it
+- `Incorrect API key provided` or another provider `AuthenticationError` means a provider key is wrong. Fix it in `.env` and restart, since the proxy reads the environment only at startup. On uv, stop it and rerun step 7. On Docker, `docker rm -f litellm` and rerun step 7, since `docker restart` keeps the environment the container was created with.
 - `Cannot automatically baseline this database` means `DATABASE_URL` points at a database that already holds tables LiteLLM did not create. Point it at a new, empty database
-- `LLM Provider NOT provided` means a `model` in `config.yaml` (or step 3) lacks its provider in front. Use the name as the step 5 lookup prints it
-- `Invalid model name passed in model=...` means the request's `model` is not a `model_name` in `config.yaml` or step 3. `GET /v1/models` lists the ones the key can use
+- `LLM Provider NOT provided` means a `model` in `config.yaml` lacks its provider in front. Use the name as the step 5 lookup prints it
+- `Invalid model name passed in model=...` means the request's `model` is not a `model_name` in `config.yaml`. `GET /v1/models` lists the ones the key can use
 - `No connected db.` on a proxy without Postgres means the bearer token is not the master key
 - `Conflict. The container name "/litellm" is already in use` means an earlier run left a container. `docker start litellm` reuses it, or with the user's OK `docker rm -f litellm` and rerun step 7
 - Inside Docker, `localhost` in `config.yaml` (an Ollama `api_base`, say) is the container itself. Use `host.docker.internal`, and on Linux also pass `--add-host=host.docker.internal:host-gateway`

.github/CODEOWNERS

@@ -1,6 +1,6 @@
 /model_prices_and_context_window.json @mateo-berri @ryan-crabbe-berri @kerry-berri @joshua-berri
 /litellm/model_prices_and_context_window_backup.json @mateo-berri @ryan-crabbe-berri @kerry-berri @joshua-berri
-/scripts/ci_cd/cost_map_pins.json @mateo-berri @ryan-crabbe-berri @kerry-berri @joshua-berri
+/litellm/scripts/ci_cd/cost_map_pins.json @mateo-berri @ryan-crabbe-berri @kerry-berri @joshua-berri
 /litellm/deploy/helm/ @yassin-berriai
 /litellm/deploy/terraform/ @yassin-berriai
 /.github/workflows/zizmor.yml @yuneng-berri

.github/CONTRIBUTING.md

@@ -360,7 +360,7 @@ Local lint and budget checks follow origin's current default branch. They refres
 ## Submitting Your PR
 
 1. **Push your branch**: `git push origin your-feature-branch`
-2. **Create a PR**: Go to GitHub and open a pull request against the repository's current default branch. Run `python3 scripts/default_branch.py --branch` to check its name
+2. **Create a PR**: Go to GitHub and open a pull request against the repository's current default branch. Run `python3 litellm/scripts/default_branch.py --branch` to check its name
 3. **Fill out the PR template**: Provide clear description of changes
 4. **Wait for review**: Maintainers will review and provide feedback
 5. **Address feedback**: Make requested changes and push updates

.github/scripts/ci/classify_changes.sh

@@ -19,7 +19,7 @@ while IFS= read -r file || [ -n "$file" ]; do
   esac
   case "$file" in
     *.md | *.mdx) : ;;
-    pyproject.toml | */pyproject.toml | uv.lock | uv.toml | .python-version | rust-toolchain.toml | litellm/rust/* | litellm/__init__.py | litellm/proxy/proxy_server.py | litellm/*mcp* | tests/*mcp* | litellm/integrations/arize/* | tests/base_sdk_tests/* | scripts/check_mcp_sdk_install.py | .github/workflows/test-mcp-dependency-resolution.yml | .github/actions/detect-changes/* | .github/actions/setup-uv-with-retries/* | .github/actions/cache-cargo-build/* | .github/scripts/detect_changes.sh | .github/scripts/uv_sync_with_retries.sh | .github/scripts/ci/classify_changes.sh | tests/unit/test_ci_path_filter.py | tests/unit/test_detect_changes.py)
+    pyproject.toml | */pyproject.toml | uv.lock | uv.toml | .python-version | rust-toolchain.toml | litellm/rust/* | litellm/__init__.py | litellm/proxy/proxy_server.py | litellm/*mcp* | tests/*mcp* | litellm/integrations/arize/* | tests/base_sdk_tests/* | litellm/scripts/check_mcp_sdk_install.py | .github/workflows/test-mcp-dependency-resolution.yml | .github/actions/detect-changes/* | .github/actions/setup-uv-with-retries/* | .github/actions/cache-cargo-build/* | .github/scripts/detect_changes.sh | .github/scripts/uv_sync_with_retries.sh | .github/scripts/ci/classify_changes.sh | tests/unit/test_ci_path_filter.py | tests/unit/test_detect_changes.py)
       has_mcp_dependencies=true ;;
   esac
   case "$file" in

.github/scripts/verify_linux_native_wheel.py

@@ -137,6 +137,9 @@ def main(
                     "litellm/proxy-extras/",
                     "litellm/rust/",
                     "litellm/packaging/",
+                    "litellm/scripts/",
+                    "litellm/runtime/",
+                    "litellm/schema.prisma",
                 )
             )
             or any(part.endswith(".dSYM") for part in PurePosixPath(member.filename).parts)

.github/workflows/auto-close-duplicates.yml

@@ -16,8 +16,8 @@ on:
   pull_request:
     paths:
       - .github/workflows/auto-close-duplicates.yml
-      - scripts/auto-close-duplicates.ts
-      - scripts/auto-close-duplicates.test.ts
+      - litellm/scripts/auto-close-duplicates.ts
+      - litellm/scripts/auto-close-duplicates.test.ts
 
 permissions: {}
 
@@ -40,7 +40,7 @@ jobs:
           bun-version: "1.4.0"
 
       - name: Test the sweep
-        run: bun test scripts/auto-close-duplicates.test.ts
+        run: bun test litellm/scripts/auto-close-duplicates.test.ts
 
   sweep:
     if: github.event_name != 'pull_request' && github.repository == 'BerriAI/litellm'
@@ -62,7 +62,7 @@ jobs:
           bun-version: "1.4.0"
 
       - name: Close unanswered duplicates, reopen ones the reporter answered
-        run: bun run scripts/auto-close-duplicates.ts
+        run: bun run litellm/scripts/auto-close-duplicates.ts
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
           DRY_RUN: ${{ inputs.dry_run == true }}

.github/workflows/check-schema-sync.yml

@@ -3,7 +3,7 @@ name: Check Schema Sync
 on:
   pull_request:
     paths:
-      - 'schema.prisma'
+      - 'litellm/schema.prisma'
       - 'litellm/proxy/schema.prisma'
       - 'litellm/proxy-extras/litellm_proxy_extras/schema.prisma'
 
@@ -16,7 +16,7 @@ concurrency:
 
 jobs:
   check-sync:
-    name: Verify schema.prisma copies match root
+    name: Verify litellm/schema.prisma copies match the canonical schema
     runs-on: ubuntu-latest
     timeout-minutes: 5
     steps:
@@ -27,7 +27,7 @@ jobs:
 
       - name: Reject symlinked schema files
         run: |
-          for f in schema.prisma litellm/proxy/schema.prisma litellm/proxy-extras/litellm_proxy_extras/schema.prisma; do
+          for f in litellm/schema.prisma litellm/proxy/schema.prisma litellm/proxy-extras/litellm_proxy_extras/schema.prisma; do
             if [ -L "$f" ]; then
               echo "::error file=$f::$f is a symlink, which is not allowed"
               exit 1
@@ -38,24 +38,24 @@ jobs:
         run: |
           EXIT=0
 
-          diff schema.prisma litellm/proxy/schema.prisma || {
-            echo "::error file=litellm/proxy/schema.prisma::litellm/proxy/schema.prisma differs from root schema.prisma"
+          diff litellm/schema.prisma litellm/proxy/schema.prisma || {
+            echo "::error file=litellm/proxy/schema.prisma::litellm/proxy/schema.prisma differs from root litellm/schema.prisma"
             EXIT=1
           }
 
-          diff schema.prisma litellm/proxy-extras/litellm_proxy_extras/schema.prisma || {
-            echo "::error file=litellm/proxy-extras/litellm_proxy_extras/schema.prisma::litellm/proxy-extras/litellm_proxy_extras/schema.prisma differs from root schema.prisma"
+          diff litellm/schema.prisma litellm/proxy-extras/litellm_proxy_extras/schema.prisma || {
+            echo "::error file=litellm/proxy-extras/litellm_proxy_extras/schema.prisma::litellm/proxy-extras/litellm_proxy_extras/schema.prisma differs from root litellm/schema.prisma"
             EXIT=1
           }
 
           if [ "$EXIT" -ne 0 ]; then
             echo ""
             echo "Schema files are out of sync."
-            echo "The root schema.prisma is the source of truth."
+            echo "The root litellm/schema.prisma is the source of truth."
             echo ""
             echo "To fix, run from the repo root:"
-            echo "  cp schema.prisma litellm/proxy/schema.prisma"
-            echo "  cp schema.prisma litellm/proxy-extras/litellm_proxy_extras/schema.prisma"
+            echo "  cp litellm/schema.prisma litellm/proxy/schema.prisma"
+            echo "  cp litellm/schema.prisma litellm/proxy-extras/litellm_proxy_extras/schema.prisma"
             exit 1
           fi
 

.github/workflows/conventional-commits.yml

@@ -3,7 +3,7 @@ name: Conventional PR Title
 # Squash-merge replaces the merge commit subject with the PR title, so
 # enforcing Conventional Commits at the PR-title level is what actually gates
 # the commits that land on the default branch. The local commit-msg hook
-# (scripts/githooks/commit-msg) is a best-effort assist; this workflow is the gate.
+# (litellm/scripts/githooks/commit-msg) is a best-effort assist; this workflow is the gate.
 #
 # See https://www.conventionalcommits.org/en/v1.0.0/
 
@@ -28,7 +28,7 @@ jobs:
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
         with:
-          # Must mirror the type list in scripts/githooks/commit-msg.
+          # Must mirror the type list in litellm/scripts/githooks/commit-msg.
           types: |
             feat
             fix

.github/workflows/cost-map-guard.yml

@@ -49,8 +49,11 @@ jobs:
           HEAD_SHA: ${{ github.event.pull_request.head.sha }}
           HEAD_REF: ${{ github.event.pull_request.head.ref }}
         run: |
-          guard=scripts/ci_cd/cost_map_guard.py
+          guard=litellm/scripts/ci_cd/cost_map_guard.py
           # TODO: Drop the legacy path after this lands on main.
+          if [ ! -f "$guard" ]; then
+            guard=scripts/ci_cd/cost_map_guard.py
+          fi
           if [ ! -f "$guard" ]; then
             guard=ci_cd/cost_map_guard.py
           fi

.github/workflows/cost-map-pins-recheck.yml

@@ -8,7 +8,7 @@ on:
     paths:
       - model_prices_and_context_window.json
       - litellm/model_prices_and_context_window_backup.json
-      - scripts/ci_cd/cost_map_pins.json
+      - litellm/scripts/ci_cd/cost_map_pins.json
 
 permissions: {}
 
@@ -28,8 +28,8 @@ jobs:
           BEFORE_SHA: ${{ github.event.before }}
           AFTER_SHA: ${{ github.event.after }}
           FORCED: ${{ github.event.forced }}
-          COST_MAP_PATHS: '["model_prices_and_context_window.json", "litellm/model_prices_and_context_window_backup.json", "scripts/ci_cd/cost_map_pins.json"]'
-          PINS_PATHS: '["scripts/ci_cd/cost_map_pins.json"]'
+          COST_MAP_PATHS: '["model_prices_and_context_window.json", "litellm/model_prices_and_context_window_backup.json", "litellm/scripts/ci_cd/cost_map_pins.json"]'
+          PINS_PATHS: '["litellm/scripts/ci_cd/cost_map_pins.json"]'
         run: |
           if [[ "$BEFORE_SHA" =~ ^0+$ || "$FORCED" == true ]] \
             || ! gh api "repos/${GITHUB_REPOSITORY}/compare/${BEFORE_SHA}...${AFTER_SHA}?per_page=1" \

.github/workflows/duplicate_issue_check.yml

@@ -13,9 +13,9 @@ on:
       - .github/workflows/duplicate_issue_check.yml
       - .github/prompts/duplicate-issue-check.md
       - .github/prompts/duplicate-issue-check.schema.json
-      - scripts/flag-duplicate-issue.ts
-      - scripts/flag-duplicate-issue.test.ts
-      - scripts/auto-close-duplicates.ts
+      - litellm/scripts/flag-duplicate-issue.ts
+      - litellm/scripts/flag-duplicate-issue.test.ts
+      - litellm/scripts/auto-close-duplicates.ts
 
 permissions: {}
 
@@ -38,7 +38,7 @@ jobs:
           bun-version: "1.4.0"
 
       - name: Test the flag step
-        run: bun test scripts/flag-duplicate-issue.test.ts
+        run: bun test litellm/scripts/flag-duplicate-issue.test.ts
 
   classify:
     if: github.event_name != 'pull_request' && github.repository == 'BerriAI/litellm'
@@ -134,7 +134,7 @@ jobs:
           bun-version: "1.4.0"
 
       - name: Comment and label
-        run: bun run scripts/flag-duplicate-issue.ts
+        run: bun run litellm/scripts/flag-duplicate-issue.ts
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
           VERDICT: ${{ needs.classify.outputs.verdict }}

.github/workflows/image-scan.yml

@@ -10,13 +10,13 @@ on:
       - litellm/deploy/docker/Dockerfile
       - litellm/deploy/docker/Dockerfile.dockerignore
       - litellm/deploy/docker/docker-entrypoint.sh
-      - migrations/run.py
+      - litellm/runtime/migrations/run.py
       - gateway/main.py
       - gateway/launch.py
       - gateway/routes/allowlist.py
       - litellm/cli/**
-      - backend/main.py
-      - backend/routes/allowlist.py
+      - litellm/runtime/backend/main.py
+      - litellm/runtime/backend/routes/allowlist.py
       - litellm/rust/**
       - litellm/proxy/lens/**
       - litellm/proxy/prisma_migration.py

.github/workflows/issue_classifier.yml

@@ -16,12 +16,12 @@ on:
       - .github/issue-labels.json
       - .github/ISSUE_TEMPLATE/bug_report.yml
       - .github/ISSUE_TEMPLATE/feature_request.yml
-      - scripts/classify-issue.ts
-      - scripts/classify-issue.test.ts
-      - scripts/label-issue.ts
-      - scripts/label-issue.test.ts
-      - scripts/issue-labels.ts
-      - scripts/auto-close-duplicates.ts
+      - litellm/scripts/classify-issue.ts
+      - litellm/scripts/classify-issue.test.ts
+      - litellm/scripts/label-issue.ts
+      - litellm/scripts/label-issue.test.ts
+      - litellm/scripts/issue-labels.ts
+      - litellm/scripts/auto-close-duplicates.ts
 
 permissions: {}
 
@@ -49,7 +49,7 @@ jobs:
           bun-version: "1.4.0"
 
       - name: Test the gate, the validation and the label step
-        run: bun test scripts/classify-issue.test.ts scripts/label-issue.test.ts
+        run: bun test litellm/scripts/classify-issue.test.ts litellm/scripts/label-issue.test.ts
 
   classify-issue:
     # An edit to a labelled issue is dropped here; the script decides the rest against the live labels
@@ -108,7 +108,7 @@ jobs:
           ISSUE_CLASSIFIER_MODEL: ${{ vars.ISSUE_CLASSIFIER_MODEL }}
         run: |
           set -euo pipefail
-          bun run scripts/classify-issue.ts > classification.json
+          bun run litellm/scripts/classify-issue.ts > classification.json
           {
             echo 'verdict<<CLASSIFICATION'
             cat classification.json
@@ -153,7 +153,7 @@ jobs:
           bun-version: "1.4.0"
 
       - name: Replace the labels in each namespace
-        run: bun run scripts/label-issue.ts
+        run: bun run litellm/scripts/label-issue.ts
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
           VERDICT: ${{ needs.classify-issue.outputs.verdict }}

.github/workflows/issue_fixed_comment.yml

@@ -15,9 +15,9 @@ on:
   pull_request:
     paths:
       - .github/workflows/issue_fixed_comment.yml
-      - scripts/comment-fixed-issue.ts
-      - scripts/comment-fixed-issue.test.ts
-      - scripts/auto-close-duplicates.ts
+      - litellm/scripts/comment-fixed-issue.ts
+      - litellm/scripts/comment-fixed-issue.test.ts
+      - litellm/scripts/auto-close-duplicates.ts
 
 permissions: {}
 
@@ -44,7 +44,7 @@ jobs:
           bun-version: "1.4.0"
 
       - name: Test the closer lookup, the release placement, the comment and the superseded pull request close
-        run: bun test scripts/comment-fixed-issue.test.ts
+        run: bun test litellm/scripts/comment-fixed-issue.test.ts
 
   comment-fixed-issue:
     if: github.event_name != 'pull_request' && github.repository == 'BerriAI/litellm'
@@ -69,7 +69,7 @@ jobs:
 
       - name: Name the release that carries the fix and close the pull requests it supersedes
         shell: bash
-        run: bun run scripts/comment-fixed-issue.ts | tee -a "${GITHUB_STEP_SUMMARY}"
+        run: bun run litellm/scripts/comment-fixed-issue.ts | tee -a "${GITHUB_STEP_SUMMARY}"
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
           ISSUE_NUMBER: ${{ github.event.issue.number || github.event.inputs.issue_number }}

.github/workflows/issue_label_sync.yml

@@ -5,7 +5,7 @@ on:
     branches: [main]
     paths:
       - .github/issue-labels.json
-      - scripts/sync-issue-labels.ts
+      - litellm/scripts/sync-issue-labels.ts
   workflow_dispatch:
     inputs:
       dry_run:
@@ -16,9 +16,9 @@ on:
     paths:
       - .github/workflows/issue_label_sync.yml
       - .github/issue-labels.json
-      - scripts/sync-issue-labels.ts
-      - scripts/sync-issue-labels.test.ts
-      - scripts/issue-labels.ts
+      - litellm/scripts/sync-issue-labels.ts
+      - litellm/scripts/sync-issue-labels.test.ts
+      - litellm/scripts/issue-labels.ts
 
 permissions: {}
 
@@ -41,7 +41,7 @@ jobs:
           bun-version: "1.4.0"
 
       - name: Test the sync
-        run: bun test scripts/sync-issue-labels.test.ts
+        run: bun test litellm/scripts/sync-issue-labels.test.ts
 
   sync-issue-labels:
     if: github.event_name != 'pull_request' && github.repository == 'BerriAI/litellm'
@@ -66,7 +66,7 @@ jobs:
           bun-version: "1.4.0"
 
       - name: Create or recolour every label in .github/issue-labels.json
-        run: bun run scripts/sync-issue-labels.ts
+        run: bun run litellm/scripts/sync-issue-labels.ts
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
           DRY_RUN: ${{ github.event_name == 'workflow_dispatch' && inputs.dry_run == true }}

.github/workflows/mutation-test.yml

@@ -129,7 +129,7 @@ jobs:
           set +e
           uv run --no-sync --with mutmut==3.5.0 mutmut export-cicd-stats > /dev/null 2>&1
           uv run --no-sync --with mutmut==3.5.0 mutmut results > mutmut-results.txt 2>&1
-          uv run --no-sync python scripts/mutation_report.py
+          uv run --no-sync python litellm/scripts/mutation_report.py
           # The full report can be very long for big test files; the run-page
           # summary cuts off at 1 MB. Append the head of the report (summary
           # + survivor list) and link out to the artifact for the full body.

.github/workflows/publish-lint-base-counts.yml

@@ -3,7 +3,7 @@ name: Publish lint base counts
 # Every commit on main can become a future merge-base.
 # Publishing its per-rule counts for each lint gate (strict ruff, type discipline,
 # test quality, basedpyright) as an artifact lets the gates download them through
-# scripts/lint_base_counts.py in seconds instead of scanning the merge-base tree in
+# litellm/scripts/lint_base_counts.py in seconds instead of scanning the merge-base tree in
 # a throwaway worktree on every fresh checkout or moved merge-base.
 # No concurrency group on purpose: runs must never cancel each other, because
 # every sha's artifact matters (any of them can become a merge-base).
@@ -30,15 +30,15 @@ jobs:
       matrix:
         include:
           - checker: ruff-strict
-            script: scripts/ruff_strict_gate.py
+            script: litellm/scripts/ruff_strict_gate.py
           - checker: type-discipline
-            script: scripts/type_discipline_gate.py
+            script: litellm/scripts/type_discipline_gate.py
           - checker: test-quality
-            script: scripts/test_quality_gate.py
+            script: litellm/scripts/test_quality_gate.py
           - checker: openapi-docs
-            script: scripts/openapi_docs_gate.py
+            script: litellm/scripts/openapi_docs_gate.py
           - checker: basedpyright
-            script: scripts/type_check_gate.py
+            script: litellm/scripts/type_check_gate.py
 
     steps:
       - uses: actions/checkout@08eba0b27e820071cde6df949e0beb9ba4906955 # v4.3.0
@@ -91,7 +91,7 @@ jobs:
       - name: Emit basedpyright counts for HEAD
         if: matrix.checker == 'basedpyright'
         run: |
-          python scripts/type_check_gate.py --emit-counts-dir "$RUNNER_TEMP/lint-counts"
+          python litellm/scripts/type_check_gate.py --emit-counts-dir "$RUNNER_TEMP/lint-counts"
 
       - name: Name the counts artifact
         run: |

.github/workflows/spend-logs-guard.yml

@@ -42,8 +42,11 @@ jobs:
           MERGE_BASE: ${{ steps.revisions.outputs.merge_base }}
           HEAD_SHA: ${{ github.event.pull_request.head.sha }}
         run: |
-          guard=scripts/ci_cd/spend_logs_guard.py
+          guard=litellm/scripts/ci_cd/spend_logs_guard.py
           # TODO: Drop the legacy path after this lands on main.
+          if [ ! -f "$guard" ]; then
+            guard=scripts/ci_cd/spend_logs_guard.py
+          fi
           if [ ! -f "$guard" ]; then
             guard=ci_cd/spend_logs_guard.py
           fi

.github/workflows/sync-schema.yml

@@ -1,9 +1,9 @@
-name: Sync schema.prisma copies
+name: Sync litellm/schema.prisma copies
 
 on:
   pull_request:
     paths:
-      - 'schema.prisma'
+      - 'litellm/schema.prisma'
 
 # Scoped to ONLY the permissions needed:
 # - contents:write to push the sync commit to the PR branch
@@ -37,7 +37,7 @@ jobs:
 
       - name: Reject symlinked schema files
         run: |
-          for f in schema.prisma litellm/proxy/schema.prisma litellm/proxy-extras/litellm_proxy_extras/schema.prisma; do
+          for f in litellm/schema.prisma litellm/proxy/schema.prisma litellm/proxy-extras/litellm_proxy_extras/schema.prisma; do
             if [ -L "$f" ]; then
               echo "::error file=$f::$f is a symlink, which is not allowed"
               exit 1
@@ -46,8 +46,8 @@ jobs:
 
       - name: Copy root schema to other locations
         run: |
-          cp schema.prisma litellm/proxy/schema.prisma
-          cp schema.prisma litellm/proxy-extras/litellm_proxy_extras/schema.prisma
+          cp litellm/schema.prisma litellm/proxy/schema.prisma
+          cp litellm/schema.prisma litellm/proxy-extras/litellm_proxy_extras/schema.prisma
 
       - name: Check for changes
         id: diff
@@ -69,5 +69,5 @@ jobs:
           git config user.email "41898282+github-actions[bot]@users.noreply.github.com"
           git checkout -B "$GITHUB_HEAD_REF"
           git add -- litellm/proxy/schema.prisma litellm/proxy-extras/litellm_proxy_extras/schema.prisma
-          git commit -m "chore: sync schema.prisma copies from root"
+          git commit -m "chore: sync litellm/schema.prisma copies from the canonical schema"
           git push origin "HEAD:$GITHUB_HEAD_REF"

.github/workflows/test-linting.yml

@@ -114,7 +114,7 @@ jobs:
       - name: Check ruff format
         if: steps.changes.outputs.decision != 'skip'
         run: |
-          git diff --name-only --diff-filter=ACMR "$GATE_BASE_SHA" HEAD -- ':(glob)litellm/**/*.py' | grep -Ev '^litellm/(enterprise|ui|vscode-extension|rust|proxy-extras|packaging|cli)/' > "$RUNNER_TEMP/ruff_format_files.txt" || true
+          git diff --name-only --diff-filter=ACMR "$GATE_BASE_SHA" HEAD -- ':(glob)litellm/**/*.py' | grep -Ev '^litellm/(enterprise|ui|vscode-extension|rust|proxy-extras|packaging|cli|scripts|runtime)/' > "$RUNNER_TEMP/ruff_format_files.txt" || true
           if [ ! -s "$RUNNER_TEMP/ruff_format_files.txt" ]; then
             echo "No changed litellm Python files to check with ruff format."
             exit 0
@@ -133,7 +133,7 @@ jobs:
 
       - name: Check MCP operation boundary
         if: steps.changes.outputs.decision != 'skip'
-        run: uv run --no-sync python scripts/check_mcp_operation_boundary.py
+        run: uv run --no-sync python litellm/scripts/check_mcp_operation_boundary.py
 
       - name: Run Ruff linting
         if: steps.changes.outputs.decision != 'skip'
@@ -152,21 +152,21 @@ jobs:
         env:
           GH_TOKEN: ${{ github.token }}
         run: |
-          uv run --no-sync python scripts/ruff_strict_gate.py --base "$GATE_BASE_SHA"
+          uv run --no-sync python litellm/scripts/ruff_strict_gate.py --base "$GATE_BASE_SHA"
 
       - name: Check type discipline (mutable collections / casts / type guards / kwargs / unexplained suppressions, delta vs merge-base counts)
         if: steps.changes.outputs.decision != 'skip'
         env:
           GH_TOKEN: ${{ github.token }}
         run: |
-          uv run --no-sync python scripts/type_discipline_gate.py --base "$GATE_BASE_SHA"
+          uv run --no-sync python litellm/scripts/type_discipline_gate.py --base "$GATE_BASE_SHA"
 
       - name: Check test quality (zero-assert / mock-echo tests, sys.path.insert, raw env writes, litellm global mutation, credential-gated skips, conftest snapshot inventory, delta vs merge-base counts)
         if: steps.changes.outputs.decision != 'skip'
         env:
           GH_TOKEN: ${{ github.token }}
         run: |
-          uv run --no-sync python scripts/test_quality_gate.py --base "$GATE_BASE_SHA"
+          uv run --no-sync python litellm/scripts/test_quality_gate.py --base "$GATE_BASE_SHA"
 
       - name: Check OpenAPI docs (POST routes without a request schema, description or 200 schema, delta vs merge-base counts)
         if: steps.changes.outputs.decision != 'skip'
@@ -174,7 +174,7 @@ jobs:
           GH_TOKEN: ${{ github.token }}
         run: |
           uv sync --frozen --group proxy-dev --group e2e-dev --extra proxy
-          uv run --no-sync python scripts/openapi_docs_gate.py --base "$GATE_BASE_SHA"
+          uv run --no-sync python litellm/scripts/openapi_docs_gate.py --base "$GATE_BASE_SHA"
 
       - name: Print OpenAI version
         if: steps.changes.outputs.decision != 'skip'
@@ -186,7 +186,7 @@ jobs:
         env:
           GH_TOKEN: ${{ github.token }}
         run: |
-          uv run --no-sync python scripts/type_check_gate.py --base "$GATE_BASE_SHA"
+          uv run --no-sync python litellm/scripts/type_check_gate.py --base "$GATE_BASE_SHA"
 
       - name: Check tests/e2e basedpyright (zero errors)
         if: steps.changes.outputs.decision != 'skip'
@@ -375,7 +375,7 @@ jobs:
             exit 0
           fi
           files="$(git diff --name-only "$base" HEAD)"
-          if grep -Ev '^litellm/(ui|vscode-extension|rust|proxy-extras|packaging|cli)/' <<< "$files" | grep -E '^(litellm/|enterprise/|(pyproject\.toml|uv\.lock)$|\.github/workflows/test-linting\.yml$)' >/dev/null || grep -Eq '^litellm/ui/litellm-dashboard/(src/lib/http/schema\.d\.ts|scripts/gen-api-types\.mjs|package(-lock)?\.json)$' <<< "$files"; then
+          if grep -Ev '^litellm/(ui|vscode-extension|rust|proxy-extras|packaging|cli|scripts|runtime)/' <<< "$files" | grep -E '^(litellm/|enterprise/|(pyproject\.toml|uv\.lock)$|\.github/workflows/test-linting\.yml$)' >/dev/null || grep -Eq '^litellm/ui/litellm-dashboard/(src/lib/http/schema\.d\.ts|scripts/gen-api-types\.mjs|package(-lock)?\.json)$' <<< "$files"; then
             echo "relevant=true" >> "$GITHUB_OUTPUT"
           else
             echo "No litellm, enterprise, dependency or generator changes in this pull request, nothing to verify."
@@ -524,7 +524,7 @@ jobs:
         run: uv run --no-sync python ./tests/code_coverage_tests/check_provider_folders_documented.py
 
       - name: check_provider_endpoints_support_generated
-        run: uv run --no-sync python scripts/generate_provider_endpoints_support.py --check
+        run: uv run --no-sync python litellm/scripts/generate_provider_endpoints_support.py --check
 
       - name: test_generate_provider_endpoints_support
         run: uv run --no-sync pytest -q --noconftest -p no:cacheprovider -c /dev/null tests/code_coverage_tests/test_generate_provider_endpoints_support.py

.github/workflows/test-mcp-dependency-resolution.yml

@@ -15,12 +15,12 @@ on:
       - "litellm/__init__.py"
       - "litellm/_version.py"
       - "litellm/_lazy_imports_registry.py"
-      - "scripts/build_core_distribution.py"
+      - "litellm/scripts/build_core_distribution.py"
       - "litellm/proxy/proxy_server.py"
       - "litellm/**/*mcp*"
       - "litellm/**/*mcp*/**"
       - "litellm/integrations/arize/**"
-      - "scripts/check_mcp_sdk_install.py"
+      - "litellm/scripts/check_mcp_sdk_install.py"
       - "tests/base_sdk_tests/**"
       - ".github/workflows/test-mcp-dependency-resolution.yml"
       - ".github/actions/detect-changes/**"
@@ -91,7 +91,7 @@ jobs:
               checker=("$GITHUB_WORKSPACE/tests/base_sdk_tests/check_base_sdk_install.py")
               (cd "$RUNNER_TEMP" && "$GITHUB_WORKSPACE/.venv-$extra/bin/python" -I "$GITHUB_WORKSPACE/tests/base_sdk_tests/check_sdk_http.py")
             else
-              checker=("$GITHUB_WORKSPACE/scripts/check_mcp_sdk_install.py" --extra "$extra")
+              checker=("$GITHUB_WORKSPACE/litellm/scripts/check_mcp_sdk_install.py" --extra "$extra")
             fi
             (cd "$RUNNER_TEMP" && "$GITHUB_WORKSPACE/.venv-$extra/bin/python" "${checker[@]}")
           done
@@ -115,7 +115,7 @@ jobs:
             if [ "$extra" = core ]; then
               checker=("$GITHUB_WORKSPACE/tests/base_sdk_tests/check_base_sdk_install.py")
             else
-              checker=("$GITHUB_WORKSPACE/scripts/check_mcp_sdk_install.py" --extra "$extra")
+              checker=("$GITHUB_WORKSPACE/litellm/scripts/check_mcp_sdk_install.py" --extra "$extra")
             fi
             (cd "$RUNNER_TEMP" && "$GITHUB_WORKSPACE/.venv-lowest-$extra/bin/python" "${checker[@]}")
           done
@@ -147,15 +147,15 @@ jobs:
       - name: Build the core wheel and source distribution
         run: >-
           uv run --no-project --with coverage==7.14.0 python -m coverage run --rcfile=/dev/null
-          --branch --include="*/scripts/build_core_distribution.py"
-          scripts/build_core_distribution.py --out-dir dist/core
+          --branch --include="*/litellm/scripts/build_core_distribution.py"
+          litellm/scripts/build_core_distribution.py --out-dir dist/core
       - name: Verify metadata, resources, and independent source rebuild
         env:
           CORE_DISTRIBUTION_DIR: ${{ github.workspace }}/dist/core
         run: >-
           uv run --no-project --with pytest==9.0.3 --with pytest-cov==5.0.0 --with coverage==7.14.0 --with 'tomli==2.4.1; python_version < "3.11"'
           python -m pytest tests/base_sdk_tests/test_core_distribution.py -v
-          --cov-config=/dev/null --cov=scripts.build_core_distribution --cov-branch --cov-append --cov-report=
+          --cov-config=/dev/null --cov=litellm.scripts.build_core_distribution --cov-branch --cov-append --cov-report=
       - name: Verify isolated core installations
         run: |
           wheel=$(realpath dist/core/litellm_core-*.whl)

.github/workflows/test-merge-smoke.yml

@@ -229,7 +229,7 @@ jobs:
             "prisma==0.11.0" \
             "openai==2.24.0"
           export PATH="/tmp/litellm-pip/bin:$PATH"
-          /tmp/litellm-pip/bin/prisma generate --schema=schema.prisma
+          /tmp/litellm-pip/bin/prisma generate --schema=litellm/schema.prisma
 
       - name: Start the fake OpenAI endpoint
         run: |

.github/workflows/test-rust.yml

@@ -6,9 +6,9 @@ on:
       - "litellm/rust/**"
       - "model_prices_and_context_window.json"
       - "litellm/rust_bridge/**"
-      - "scripts/generate_trace_types.py"
-      - "scripts/trace_codegen/**"
-      - "scripts/lens_assets/**"
+      - "litellm/scripts/generate_trace_types.py"
+      - "litellm/scripts/trace_codegen/**"
+      - "litellm/scripts/lens_assets/**"
       - "litellm/tracing/**"
       - "tests/test_litellm_rust/**"
       - "litellm/integrations/custom_logger.py"
@@ -41,9 +41,9 @@ on:
       - "litellm/rust/**"
       - "model_prices_and_context_window.json"
       - "litellm/rust_bridge/**"
-      - "scripts/generate_trace_types.py"
-      - "scripts/trace_codegen/**"
-      - "scripts/lens_assets/**"
+      - "litellm/scripts/generate_trace_types.py"
+      - "litellm/scripts/trace_codegen/**"
+      - "litellm/scripts/lens_assets/**"
       - "litellm/tracing/**"
       - "tests/test_litellm_rust/**"
       - "litellm/integrations/custom_logger.py"
@@ -160,7 +160,7 @@ jobs:
 
       - name: Check generated trace contracts
         working-directory: .
-        run: uv run scripts/generate_trace_types.py --check
+        run: uv run litellm/scripts/generate_trace_types.py --check
 
       - run: cargo nextest run --workspace --locked
 

.github/workflows/test-unit.yml

@@ -1043,7 +1043,7 @@ jobs:
           fi
           uv run --no-sync python -c 'import os, sys; print(sys.version); assert f"{sys.version_info.major}.{sys.version_info.minor}" == os.environ["UV_PYTHON"]'
 
-      - name: Check schema.prisma against the merge base
+      - name: Check litellm/schema.prisma against the merge base
         if: steps.changes.outputs.decision != 'skip' && matrix.shard == 'db-and-spend' && github.event_name == 'pull_request'
         timeout-minutes: 3
         env:
@@ -1056,7 +1056,7 @@ jobs:
           merge_base=$(gh api "repos/${GITHUB_REPOSITORY}/compare/${base_ref}...${HEAD_SHA}?per_page=1" --jq '.merge_base_commit.sha')
           test -n "$merge_base"
           git fetch --no-tags --depth=1 origin "$merge_base"
-          uv run --no-sync python scripts/check_schema_compatibility.py --base-commit "$merge_base"
+          uv run --no-sync python litellm/scripts/check_schema_compatibility.py --base-commit "$merge_base"
 
       - name: Cache Prisma binaries
         if: steps.changes.outputs.decision != 'skip'

.gitignore

@@ -106,7 +106,7 @@ litellm_config.yaml
 litellm/proxy/to_delete_loadtest_work/*
 update_model_cost_map.py
 tests/unit/proxy/mcp_server/test_mcp_server_manager.py
-scripts/test_vertex_ai_search.py
+litellm/scripts/test_vertex_ai_search.py
 LAZY_LOADING_IMPROVEMENTS.md
 STABILIZATION_TODO.md
 **/test-results

AGENTS.md

@@ -30,7 +30,7 @@ A test must only fail when litellm code changes. Never pin facts we don't own (a
 
 End-to-end tests belong in `tests/e2e/`. A rule about one directory's files, like its harness or which CI runs it, goes in that directory's `AGENTS.md`, never only in a README or here. Read a directory's `AGENTS.md` before editing files in it
 
-When creating PRs, target the repository's current default branch for both internal and external / OSS contributions. Check it with `python3 scripts/default_branch.py --branch` instead of assuming a branch name or relying on cached `origin/HEAD`
+When creating PRs, target the repository's current default branch for both internal and external / OSS contributions. Check it with `python3 litellm/scripts/default_branch.py --branch` instead of assuming a branch name or relying on cached `origin/HEAD`
 
 When writing a PR body, treat the comments and imperative instructions inside .github/PULL_REQUEST_TEMPLATE/general.md as rules to follow, not just layout. Agent harnesses may strip HTML comments from copies of that file injected into context, so read .github/PULL_REQUEST_TEMPLATE/general.md from disk before writing a PR body to make sure you see every comment rule. A section you have nothing to put in (Relevant issues, Affected release, Linear ticket, Caveats, QA runbook, and so on) is removed entirely, heading included, never left as an empty title
 
@@ -56,11 +56,11 @@ Don't hesitate to use values in .env to get needed API keys and other secrets, a
 
 Python max line length is 120, not 88
 
-The five lint gates (`scripts/ruff_strict_gate.py`, `scripts/type_discipline_gate.py`, `scripts/test_quality_gate.py`, `scripts/openapi_docs_gate.py`, `scripts/type_check_gate.py`) compare each rule's codebase count on your branch against the count at its merge-base with the default branch, and a rule may not grow. There are no budget files to edit or ratchet: when a gate fails, fix the violations the branch introduced or remove at least as many of that rule elsewhere in the tree. The one exception is `reportAny` / `reportExplicitAny`, which share a fixed codebase-wide cap in `ANY_CAPS` in `scripts/type_check_gate.py` because Any spreads past the lines you touch. A branch may add Anys while the total stays under that cap. Only lower `ANY_CAPS`, and only in its own PR to the default branch, never raise it on a feature branch
+The five lint gates (`litellm/scripts/ruff_strict_gate.py`, `litellm/scripts/type_discipline_gate.py`, `litellm/scripts/test_quality_gate.py`, `litellm/scripts/openapi_docs_gate.py`, `litellm/scripts/type_check_gate.py`) compare each rule's codebase count on your branch against the count at its merge-base with the default branch, and a rule may not grow. There are no budget files to edit or ratchet: when a gate fails, fix the violations the branch introduced or remove at least as many of that rule elsewhere in the tree. The one exception is `reportAny` / `reportExplicitAny`, which share a fixed codebase-wide cap in `ANY_CAPS` in `litellm/scripts/type_check_gate.py` because Any spreads past the lines you touch. A branch may add Anys while the total stays under that cap. Only lower `ANY_CAPS`, and only in its own PR to the default branch, never raise it on a feature branch
 
 `make check` (f.k.a. `make pre-commit`, which still works identically as an alias) saves its complete output to a log file in .git (overwriting previous logs) and prints that path as its first and last output lines. To inspect a run, read or grep that log instead of re-running the multi-minute checks just to see a different slice
 
-`make check`, `make lint`, `scripts/pre_commit_lint.sh`, and the standalone lint gates (`scripts/ruff_strict_gate.py`, `scripts/type_discipline_gate.py`, `scripts/type_check_gate.py`) each hold one of 2 machine-wide slots, so when other sessions or worktrees on the same box are already running heavy work, yours prints "all N machine-wide slots are busy" followed by "queueing" and then stays quiet until a slot frees. Give the command a long timeout and let it wait rather than killing it, retrying it, or assuming it hung. Don't change the # of machine-wide slots or make it unlimited by setting `LITELLM_GATE_SLOTS=0`
+`make check`, `make lint`, `litellm/scripts/pre_commit_lint.sh`, and the standalone lint gates (`litellm/scripts/ruff_strict_gate.py`, `litellm/scripts/type_discipline_gate.py`, `litellm/scripts/type_check_gate.py`) each hold one of 2 machine-wide slots, so when other sessions or worktrees on the same box are already running heavy work, yours prints "all N machine-wide slots are busy" followed by "queueing" and then stays quiet until a slot frees. Give the command a long timeout and let it wait rather than killing it, retrying it, or assuming it hung. Don't change the # of machine-wide slots or make it unlimited by setting `LITELLM_GATE_SLOTS=0`
 
 If you're trying to create a new function that relies on untyped stuff, instead of adding more Any's and pushing `reportAny` / `reportExplicitAny` closer to their cap in `ANY_CAPS`, just validate it in the caller with Pydantic (a model or `TypeAdapter` that returns the typed thing or raises will do) and then pass the now typed variable in
 

Makefile

@@ -63,11 +63,11 @@ UV := uv
 UV_RUN := $(UV) run --no-sync
 BASE_REF ?=
 export BASE_REF
-RESOLVE_BASE = python3 scripts/default_branch.py --base "$(BASE_REF)"
+RESOLVE_BASE = python3 litellm/scripts/default_branch.py --base "$(BASE_REF)"
 
 # Machine-wide slot queue for the heavy targets below; python3 + stdlib only, so
-# it runs before any venv exists. See scripts/gate_slot_lock.py.
-GATE_SLOT_LOCK := python3 scripts/gate_slot_lock.py
+# it runs before any venv exists. See litellm/scripts/gate_slot_lock.py.
+GATE_SLOT_LOCK := python3 litellm/scripts/gate_slot_lock.py
 
 LINT_DEP_INSTALL ?= install-dev
 LINT_E2E_DEP_INSTALL ?= lint-install
@@ -90,8 +90,8 @@ install-dev:
 # machine-wide slots the CPU-bound gates below share.
 bootstrap:
 	$(UV) sync --inexact --frozen --extra proxy --group proxy-dev --group e2e-dev
-	$(UV_RUN) python scripts/prisma_generate_if_needed.py
-	cd litellm/ui/litellm-dashboard && ../../../scripts/with_dashboard_node.sh npm install --no-audit --no-fund
+	$(UV_RUN) python litellm/scripts/prisma_generate_if_needed.py
+	cd litellm/ui/litellm-dashboard && ../../scripts/with_dashboard_node.sh npm install --no-audit --no-fund
 	@main_root=$$(git worktree list --porcelain | head -1 | sed 's/^worktree //'); \
 	if [ "$$main_root" != "$$(git rev-parse --show-toplevel)" ] && [ -f "$$main_root/.env" ] && [ ! -f .env ]; then \
 		cp "$$main_root/.env" .env && echo "bootstrap: copied .env from $$main_root"; \
@@ -123,7 +123,7 @@ install-helm-unittest:
 # Install git hooks that enforce Conventional Commits.
 # Opt-in: not chained into install-dev.
 install-hooks:
-	./scripts/install_git_hooks.sh
+	./litellm/scripts/install_git_hooks.sh
 
 # Formatting
 # Wrap width is pyproject.toml's single source of truth (line-length = 120), shared by the
@@ -139,12 +139,12 @@ lint-fetch-base:
 
 # Mirror test-linting.yml's python job environment: the proxy-dev group plus a generated
 # Prisma client, so `basedpyright tests/e2e` resolves the same modules CI does. The
-# basedpyright gate itself no longer measures here (scripts/type_check_gate.py provisions its
+# basedpyright gate itself no longer measures here (litellm/scripts/type_check_gate.py provisions its
 # own .venv-typecheck). --inexact tops up the venv instead of pruning the proxy extras
 # gen:api and the running proxy need.
 lint-install:
 	$(UV) sync --inexact --frozen --group proxy-dev --group e2e-dev
-	$(UV_RUN) python scripts/prisma_generate_if_needed.py
+	$(UV_RUN) python litellm/scripts/prisma_generate_if_needed.py
 
 # Diff-scoped format check, mirroring test-linting.yml's "Check ruff format" step:
 # only the litellm Python files changed vs the base are checked, so a pre-existing
@@ -163,7 +163,7 @@ lint-format-check-changed: $(LINT_DEP_INSTALL) $(LINT_DEP_BASE)
 
 # Linting targets
 lint-ruff: $(LINT_DEP_INSTALL)
-	$(UV_RUN) python scripts/check_mcp_operation_boundary.py
+	$(UV_RUN) python litellm/scripts/check_mcp_operation_boundary.py
 	cd litellm && $(UV_RUN) ruff check . && cd ..
 	$(UV_RUN) ruff check --config .github/linters/ruff-tests.toml tests
 
@@ -204,7 +204,7 @@ lint-ruff-FULL-dev: install-dev
 	else echo "No changed .py files to check."; fi
 
 lint-basedpyright: $(LINT_DEP_INSTALL) $(LINT_DEP_BASE)
-	$(UV_RUN) python scripts/type_check_gate.py --base "$(BASE_REF)"
+	$(UV_RUN) python litellm/scripts/type_check_gate.py --base "$(BASE_REF)"
 
 lint-e2e-basedpyright: $(LINT_E2E_DEP_INSTALL)
 	$(UV_RUN) basedpyright tests/e2e tests/e2e_harness
@@ -212,29 +212,29 @@ lint-e2e-basedpyright: $(LINT_E2E_DEP_INSTALL)
 # Type-discipline gate (mutable collections / casts / type guards / kwargs /
 # unexplained suppressions), the test-linting.yml step `make lint` used to omit.
 lint-type-discipline: $(LINT_DEP_INSTALL) $(LINT_DEP_BASE)
-	$(UV_RUN) python scripts/type_discipline_gate.py --base "$(BASE_REF)"
+	$(UV_RUN) python litellm/scripts/type_discipline_gate.py --base "$(BASE_REF)"
 
 # Test-quality gate (zero-assert / mock-echo tests, sys.path.insert, raw env writes,
 # litellm module-global mutation, credential-gated skips, conftest snapshot
 # inventory), counted across tests/ the same delta-vs-base way.
 lint-test-quality: $(LINT_DEP_INSTALL) $(LINT_DEP_BASE)
-	$(UV_RUN) python scripts/test_quality_gate.py --base "$(BASE_REF)"
+	$(UV_RUN) python litellm/scripts/test_quality_gate.py --base "$(BASE_REF)"
 
 # OpenAPI docs gate (POST routes without a request body schema, a description or a
 # 200 schema in app.openapi()), counted the same delta-vs-base way so a new endpoint
 # has to show up documented in Swagger UI.
 lint-openapi-docs: $(LINT_DEP_INSTALL) $(LINT_DEP_BASE)
-	$(UV_RUN) python scripts/openapi_docs_gate.py --base "$(BASE_REF)"
+	$(UV_RUN) python litellm/scripts/openapi_docs_gate.py --base "$(BASE_REF)"
 
 lint-format: format-check
 
 lint-ruff-strict: install-dev
-	$(UV_RUN) python scripts/ruff_strict_gate.py --base "$(BASE_REF)"
+	$(UV_RUN) python litellm/scripts/ruff_strict_gate.py --base "$(BASE_REF)"
 
 # Strict gate, invoked the same way CI does in test-linting.yml so a local pass
 # means the CI check will pass too.
 lint-gate: $(LINT_DEP_INSTALL) $(LINT_DEP_BASE)
-	$(UV_RUN) python scripts/ruff_strict_gate.py --base "$(BASE_REF)"
+	$(UV_RUN) python litellm/scripts/ruff_strict_gate.py --base "$(BASE_REF)"
 
 check-circular-imports: $(LINT_DEP_INSTALL)
 	cd litellm && $(UV_RUN) python ../tests/documentation_tests/test_circular_imports.py && cd ..
@@ -272,7 +272,7 @@ check:
 	@$(GATE_SLOT_LOCK) $(MAKE) check-inner
 
 check-inner: bootstrap
-	./scripts/pre_commit_lint.sh
+	./litellm/scripts/pre_commit_lint.sh
 
 pre-commit:
 	@echo "make pre-commit is a legacy alias; use make check" >&2

README.md

@@ -99,7 +99,7 @@ proxy, CLI, and optional extras
 Build and install core from this checkout while its release integration is pending:
 
 ```shell
-python scripts/build_core_distribution.py --out-dir dist/core
+python litellm/scripts/build_core_distribution.py --out-dir dist/core
 python -m pip install dist/core/litellm_core-*.whl
 ```
 

litellm/deploy/docker/Dockerfile

@@ -115,11 +115,11 @@ RUN uv sync --frozen --no-install-package litellm-proxy --no-default-groups --no
 
 RUN uv pip install --python /app/.venv/bin/python --no-deps /wheels/litellm_proxy-*.whl
 
-RUN cp -r litellm/enterprise "$(.venv/bin/python -c 'import sysconfig; print(sysconfig.get_path("purelib"))')/litellm/enterprise"
+RUN cp -r litellm/enterprise litellm/runtime "$(.venv/bin/python -c 'import sysconfig; print(sysconfig.get_path("purelib"))')/litellm/"
 
 RUN HOME=/opt/prisma XDG_CACHE_HOME=/opt/prisma/.cache PRISMA_BINARY_CACHE_DIR=/opt/prisma/binaries \
     npm_config_cache=/root/.npm \
-    prisma generate --schema=./schema.prisma
+    prisma generate --schema=litellm/schema.prisma
 
 RUN sed -i 's/\r$//' litellm/deploy/docker/docker-entrypoint.sh && chmod +x litellm/deploy/docker/docker-entrypoint.sh
 
@@ -180,14 +180,12 @@ ENV PATH="/app/.venv/bin:${PATH}" \
 COPY --from=builder /app/.venv /app/.venv
 COPY --from=liteadmin-builder /opt/liteadmin /opt/liteadmin
 COPY --from=builder /app/litellm/deploy/docker/docker-entrypoint.sh /app/docker-entrypoint.sh
-COPY --from=builder /app/schema.prisma /app/schema.prisma
+COPY --from=builder /app/litellm/schema.prisma /app/litellm/schema.prisma
+COPY --from=builder /app/gateway /app/gateway
 COPY --from=builder /app/litellm/proxy/prisma_migration.py /app/litellm/proxy/prisma_migration.py
 # enterprise/ is imported by source path at runtime (proxy_cli puts the
 # working directory on sys.path; litellm/proxy/hooks resolves
 # litellm.enterprise.enterprise_hooks from it)
-COPY --from=builder /app/gateway /app/gateway
-COPY --from=builder /app/backend /app/backend
-COPY --from=builder /app/migrations /app/migrations
 COPY --from=builder /opt/prisma /opt/prisma
 COPY --from=builder /var/lib/litellm /var/lib/litellm
 COPY litellm/ui/nginx.conf /etc/nginx/nginx.conf
@@ -205,7 +203,7 @@ RUN find /app/.venv -depth -type d -path "*/tornado/test" -exec rm -rf {} + && \
     ls /opt/prisma/binaries/node_modules/@prisma/engines/query-engine-* >/dev/null && \
     python -c "from prisma.client import BINARY_PATHS; paths = list(BINARY_PATHS.query_engine.values()); assert paths and all(p.startswith('/opt/prisma/') for p in paths), paths" && \
     python -c "from litellm.rust_bridge.loader import native_bridge_available; assert native_bridge_available()" && \
-    python -c "import gateway.launch, backend.main"
+    python -c "import gateway.launch, litellm.runtime.backend.main"
 
 USER 65532:65532
 

litellm/deploy/docker/README.md

@@ -30,9 +30,9 @@ Every LiteLLM container runs the same image. The first word of the container com
 |--------------|-----------------------------------------------------|------|
 | `proxy`      | `litellm ...` (everything in one process, default)  | 4000 |
 | `gateway`    | `python -m gateway.launch ...` (inference routes)   | 4000 |
-| `backend`    | `uvicorn backend.main:app ...` (management routes)  | 4001 |
+| `backend`    | `uvicorn litellm.runtime.backend.main:app ...` (management routes)  | 4001 |
 | `ui`         | nginx serving the static admin UI                   | 3000 |
-| `migrations` | `python migrations/run.py`, `prisma migrate deploy` then exit | |
+| `migrations` | `python -m litellm.runtime.migrations.run`, `prisma migrate deploy` then exit | |
 | `metrics`    | `python -m litellm.proxy.prometheus_metrics_server ...` | `--port` |
 | `collector`  | `python -m litellm.proxy.collector ...`             | |
 

litellm/deploy/docker/docker-entrypoint.sh

@@ -8,9 +8,9 @@
 #
 #   proxy       litellm ARGS                     everything in one process (default)
 #   gateway     python -m gateway.launch ARGS    inference routes, 0.0.0.0:4000
-#   backend     uvicorn backend.main:app ARGS    management routes, 0.0.0.0:4001
+#   backend     uvicorn litellm.runtime.backend.main:app ARGS    management routes, 0.0.0.0:4001
 #   ui          nginx serving the admin UI       port 3000
-#   migrations  python migrations/run.py         prisma migrate deploy, then exit
+#   migrations  python litellm/runtime/migrations/run.py         prisma migrate deploy, then exit
 #   metrics     python -m litellm.proxy.prometheus_metrics_server ARGS
 #   collector   python -m litellm.proxy.collector ARGS
 #   --admin-agent  the bundled LiteAdmin Slack agent (litellm-admin-agent --web ARGS)
@@ -88,13 +88,13 @@ os.execvp("litellm", ["litellm", *sys.argv[1:]])' "$@"
         set -- python -m gateway.launch --workers "${NUM_WORKERS:-1}" --host 0.0.0.0 --port 4000 "$@"
         ;;
     backend)
-        set -- uvicorn backend.main:app --host 0.0.0.0 --port 4001 "$@"
+        set -- uvicorn litellm.runtime.backend.main:app --host 0.0.0.0 --port 4001 "$@"
         ;;
     ui)
         exec nginx -g 'daemon off;' "$@"
         ;;
     migrations)
-        set -- python /app/migrations/run.py "$@"
+        set -- python -m litellm.runtime.migrations.run "$@"
         ;;
     metrics)
         set -- python -m litellm.proxy.prometheus_metrics_server "$@"

litellm/deploy/helm/litellm/templates/_helpers.tpl

@@ -296,7 +296,7 @@ by the controller rather than declared, so nothing there can collide.
 LITELLM_MASTER_KEY, and PROXY_MASTER_KEY for configs copied from litellm-helm,
 for the app containers only. The migrations Job runs as a
 pre-install hook, before the chart's generated master key Secret exists, and
-migrations/run.py never reads the key, so the Job must not reference it.
+litellm/runtime/migrations/run.py never reads the key, so the Job must not reference it.
 */}}
 {{- define "litellm.masterKeyEnv" -}}
 {{- range list "LITELLM_MASTER_KEY" "PROXY_MASTER_KEY" }}
@@ -507,7 +507,7 @@ The migrations Job (helm.sh/hook: pre-upgrade) is the single owner of
 Prisma schema-update on startup and contends with the Job — and with each
 other — for Prisma's Postgres advisory lock on the writer, which makes the
 Job's `migrate deploy` intermittently block until its per-attempt timeout
-and retry-exhaust. The Job's entrypoint (migrations/run.py) does not import
+and retry-exhaust. The Job's entrypoint (litellm/runtime/migrations/run.py) does not import
 proxy_server and never reads DISABLE_SCHEMA_UPDATE, so emitting it here is a
 harmless no-op for the Job and authoritative for the app pods.
 */}}

litellm/packaging/litellm-core/pyproject.toml

@@ -66,6 +66,11 @@ include = [
     "litellm/proxy/common_utils/codex_bedrock_runtime_models_0.166.0.json",
 ]
 exclude = [
+    "litellm/scripts",
+    "litellm/scripts/**",
+    "litellm/runtime",
+    "litellm/runtime/**",
+    "litellm/schema.prisma",
     "litellm/enterprise",
     "litellm/enterprise/**",
     "litellm/proxy-extras",

litellm/proxy-extras/litellm_proxy_extras/request_log_indexes.py

@@ -4,7 +4,7 @@ background, once it serves).
 
 A migration cannot build them: a plain `CREATE INDEX` blocks spend-log inserts for the
 whole build, and `CREATE INDEX CONCURRENTLY` is refused on a partitioned parent
-(scripts/db/partition_spend_logs.sql). `REQUEST_LOG_INDEXES` is the one list to extend;
+(litellm/scripts/db/partition_spend_logs.sql). `REQUEST_LOG_INDEXES` is the one list to extend;
 names match what Prisma derives from the `@@index` declarations in schema.prisma, so an
 index a database already has is recognized and never rebuilt.
 """

litellm/proxy-extras/litellm_proxy_extras/utils.py

@@ -147,7 +147,7 @@ _SPEND_LOGS_PK_CLAUSE_RE = re.compile(
 )
 
 PARTITIONED_SPEND_LOGS_PUSH_ERROR = (
-    "LiteLLM_SpendLogs is a partitioned table (see scripts/db/partition_spend_logs.sql), "
+    "LiteLLM_SpendLogs is a partitioned table (see litellm/scripts/db/partition_spend_logs.sql), "
     'so its primary key must include the partition key ("startTime"). `prisma db push` '
     "reconciles the database against schema.prisma, which declares the unpartitioned "
     'primary key ("request_id"), and Postgres rejects that rewrite with: unique '
@@ -179,7 +179,7 @@ def _without_spend_logs_pk_clauses(statement: str) -> str | None:
 
 def filter_partitioned_spend_logs_diff(diff_sql: str) -> str:
     """Drop statements from a `prisma migrate diff` script that fight the
-    SpendLogs partitioning runbook (scripts/db/partition_spend_logs.sql): the
+    SpendLogs partitioning runbook (litellm/scripts/db/partition_spend_logs.sql): the
     primary-key rewrite on "LiteLLM_SpendLogs", which Postgres rejects on a
     partitioned table, and drops of runbook artifacts such as
     "LiteLLM_SpendLogs_legacy"."""
@@ -679,7 +679,7 @@ class ProxyExtrasDBManager:
         """True when the connected database's LiteLLM_SpendLogs is a
         partitioned table in Prisma's target schema (the `schema` URL param,
         falling back to Prisma's default target, public), i.e. the operator
-        ran scripts/db/partition_spend_logs.sql. Returns False when psycopg is
+        ran litellm/scripts/db/partition_spend_logs.sql. Returns False when psycopg is
         unavailable or the database cannot be reached, preserving the
         pre-existing behavior in those cases."""
         database_url = os.getenv("DATABASE_URL")
@@ -692,7 +692,7 @@ class ProxyExtrasDBManager:
             logger.warning(
                 "psycopg is not installed; skipping the LiteLLM_SpendLogs "
                 "partition check. If this table is partitioned (see "
-                "scripts/db/partition_spend_logs.sql), schema reconciliation "
+                "litellm/scripts/db/partition_spend_logs.sql), schema reconciliation "
                 "will try to rewrite its primary key and fail. Install the "
                 "litellm[extra_proxy] extra, which now includes psycopg."
             )

litellm/proxy-extras/migration_runbook.md

@@ -10,7 +10,7 @@ Before doing anything else, make sure all `schema.prisma` files in the repo are
 
 | File | Purpose |
 |------|---------|
-| `schema.prisma` (repo root) | Source of truth |
+| `litellm/schema.prisma` (canonical) | Source of truth |
 | `litellm/proxy/schema.prisma` | Used by the proxy server |
 | `litellm/proxy-extras/litellm_proxy_extras/schema.prisma` | Used for migration generation |
 
@@ -18,16 +18,16 @@ Before doing anything else, make sure all `schema.prisma` files in the repo are
 
 ```bash
 # 1. Diff all schema files against the root source of truth
-diff schema.prisma litellm/proxy/schema.prisma
-diff schema.prisma litellm/proxy-extras/litellm_proxy_extras/schema.prisma
+diff litellm/schema.prisma litellm/proxy/schema.prisma
+diff litellm/schema.prisma litellm/proxy-extras/litellm_proxy_extras/schema.prisma
 
 # 2. If there are differences, copy the root schema to all locations
-cp schema.prisma litellm/proxy/schema.prisma
-cp schema.prisma litellm/proxy-extras/litellm_proxy_extras/schema.prisma
+cp litellm/schema.prisma litellm/proxy/schema.prisma
+cp litellm/schema.prisma litellm/proxy-extras/litellm_proxy_extras/schema.prisma
 
 # 3. Verify all files are now identical
-diff schema.prisma litellm/proxy/schema.prisma && echo "proxy schema in sync" || echo "MISMATCH"
-diff schema.prisma litellm/proxy-extras/litellm_proxy_extras/schema.prisma && echo "extras schema in sync" || echo "MISMATCH"
+diff litellm/schema.prisma litellm/proxy/schema.prisma && echo "proxy schema in sync" || echo "MISMATCH"
+diff litellm/schema.prisma litellm/proxy-extras/litellm_proxy_extras/schema.prisma && echo "extras schema in sync" || echo "MISMATCH"
 ```
 
 > **Do NOT proceed to migration generation until all schema files are identical.**
@@ -43,7 +43,7 @@ brew install postgresql@14  # macOS
 export PATH="/opt/homebrew/opt/postgresql@14/bin:$PATH"
 
 # Run migration
-uv run --with testing.postgresql python scripts/ci_cd/run_migration.py "your_migration_name"
+uv run --with testing.postgresql python litellm/scripts/ci_cd/run_migration.py "your_migration_name"
 ```
 
 ## What It Does
@@ -69,7 +69,7 @@ When the guard fires:
 1. Update your branch:
 
    ```bash
-   base_branch=$(python3 scripts/default_branch.py --branch) &&
+   base_branch=$(python3 litellm/scripts/default_branch.py --branch) &&
      git fetch origin "+refs/heads/$base_branch:refs/remotes/origin/$base_branch" &&
      git rebase "origin/$base_branch"
    ```
@@ -89,7 +89,7 @@ When the guard fires:
 4. Only if the drops are genuinely intentional, re-run with the flag:
 
    ```bash
-   uv run --with testing.postgresql python scripts/ci_cd/run_migration.py "your_migration_name" --allow-destructive
+   uv run --with testing.postgresql python litellm/scripts/ci_cd/run_migration.py "your_migration_name" --allow-destructive
    ```
 
 > **AI AGENTS / ASSISTANTS:** Do **not** automatically re-run the command with `--allow-destructive`. If the guard fires while you are driving the runbook for a human, stop, show them the error, and wait for their explicit confirmation before passing the flag. Auto-passing `--allow-destructive` is the exact failure mode this guard exists to prevent.
@@ -98,7 +98,7 @@ When the guard fires:
 
 **Missing testing module:**
 ```bash
-uv run --with testing.postgresql python scripts/ci_cd/run_migration.py "your_migration_name"
+uv run --with testing.postgresql python litellm/scripts/ci_cd/run_migration.py "your_migration_name"
 ```
 
 **initdb not found:**
@@ -115,7 +115,7 @@ rm -rf litellm/proxy-extras/litellm_proxy_extras/migrations/[empty_dir]
 ## Rules
 
 - Sync all `schema.prisma` files first (Step 0)
-- Update `schema.prisma` at the repo root first, then sync copies
+- Update `litellm/schema.prisma` first, then sync copies
 - Review generated SQL before committing
 - Use descriptive migration names
 - Never edit existing migration files

litellm/proxy/_types.py

@@ -3004,7 +3004,7 @@ class ConfigGeneralSettings(LiteLLMPydanticObjectBase):
     )
     use_spend_logs_partitioning: bool | None = Field(
         None,
-        description="If True and LiteLLM_SpendLogs has been converted to a range-partitioned table (scripts/db/partition_spend_logs.sql), retention cleanup drops expired partitions instead of deleting rows, and pre-creates upcoming partitions. Default is False.",
+        description="If True and LiteLLM_SpendLogs has been converted to a range-partitioned table (litellm/scripts/db/partition_spend_logs.sql), retention cleanup drops expired partitions instead of deleting rows, and pre-creates upcoming partitions. Default is False.",
     )
     maximum_spend_logs_cleanup_batch_size: int | None = Field(
         None,

litellm/proxy/client/cli/README.md

@@ -579,7 +579,7 @@ The Codex undo receipt is kept in a private `.litellm` directory beside the reso
 `lite configure claude` wires Claude Code up persistently with a long-lived virtual key, a pinned model and an undo, and `lite unconfigure claude` puts things back:
 
 ```bash
-curl -fsSL https://raw.githubusercontent.com/BerriAI/litellm/main/scripts/install.sh | sh
+curl -fsSL https://raw.githubusercontent.com/BerriAI/litellm/main/litellm/scripts/install.sh | sh
 lite --base-url https://your-proxy.example.com configure claude --api-key sk-... --model claude-auto
 claude
 ```
@@ -615,7 +615,7 @@ After upgrading the CLI, run `lite configure claude` to refresh `~/.litellm/stat
 `lite autoroute start` builds and runs a throwaway litellm proxy locally, so unlike the rest of this CLI it needs the proxy server runtime, not just the thin `litellm[cli]` client. Install `litellm[proxy]` (which ships the `lite` command too) with a single curl command -- no existing Python tooling required, `uv` is bootstrapped automatically if missing:
 
 ```bash
-curl -fsSL https://raw.githubusercontent.com/BerriAI/litellm/main/scripts/install.sh | sh
+curl -fsSL https://raw.githubusercontent.com/BerriAI/litellm/main/litellm/scripts/install.sh | sh
 ```
 
 To QA an unreleased branch or commit instead of the latest PyPI release, set `LITELLM_CLI_REF`:
@@ -625,7 +625,7 @@ curl -fsSL https://raw.githubusercontent.com/BerriAI/litellm/<branch-or-commit>/
   LITELLM_CLI_REF=<branch-or-commit> sh
 ```
 
-The thin `scripts/install-cli.sh` installs only `litellm[cli]`, which is enough for `lite login`, `lite claude`, and `lite up`, but not for `lite autoroute start`; running it against a `litellm[cli]` install fails fast with a message telling you to install the proxy runtime.
+The thin `litellm/scripts/install-cli.sh` installs only `litellm[cli]`, which is enough for `lite login`, `lite claude`, and `lite up`, but not for `lite autoroute start`; running it against a `litellm[cli]` install fails fast with a message telling you to install the proxy runtime.
 
 Point the CLI at your real proxy and key before running any `lite model-groups` or `lite autoroute` command -- like every other command in this CLI, they read `LITELLM_PROXY_URL`/`LITELLM_PROXY_API_KEY` (or `--base-url`/`--api-key`), no `lite login` required:
 

litellm/proxy/db/create_views.py

@@ -51,7 +51,7 @@ async def create_view_tolerating_race(db: SupportsExecuteRaw, view_name: str, dd
 async def create_missing_views(db: SupportsRawQueries) -> None:
     """
     --------------------------------------------------
-    NOTE: Copy of `scripts/db/create_views.py`.
+    NOTE: Copy of `litellm/scripts/db/create_views.py`.
     --------------------------------------------------
     Checks if the LiteLLM_VerificationTokenView and MonthlyGlobalSpend exists in the user's db.
 

litellm/proxy/db/db_transaction_queue/spend_logs_partition_manager.py

@@ -9,7 +9,7 @@ to the OS immediately, with no tombstones and no vacuum.
 
 This manager only acts when use_spend_logs_partitioning is enabled in
 general_settings AND the table is already partitioned (set up via the
-scripts/db/partition_spend_logs.sql runbook). Without both, the cleanup job
+litellm/scripts/db/partition_spend_logs.sql runbook). Without both, the cleanup job
 keeps the batched-DELETE path, so existing deployments are untouched.
 """
 

litellm/runtime/backend/main.py

@@ -5,7 +5,7 @@ route table to just the management/admin surface used by the dashboard. Purely
 additive — no existing module is modified.
 
 Run with:
-    uvicorn backend.main:app --host 0.0.0.0 --port 4001
+    uvicorn litellm.runtime.backend.main:app --host 0.0.0.0 --port 4001
 """
 
 from collections.abc import AsyncGenerator, Mapping
@@ -24,7 +24,7 @@ DatabaseURLSettings.from_env().apply_to_env()
 
 from litellm.proxy.proxy_server import app
 
-from backend.routes.allowlist import (
+from litellm.runtime.backend.routes.allowlist import (
     BACKEND_EXACT_PATHS,
     BACKEND_MOUNT_PATHS,
     BACKEND_PATH_PREFIXES,

litellm/scripts/adaptive_router_demo/README.md

@@ -76,7 +76,7 @@ The dashboard is a single static HTML file. Either:
 - **If your browser blocks `file://` fetches:**
 
   ```bash
-  cd scripts/adaptive_router_demo
+  cd litellm/scripts/adaptive_router_demo
   python3 -m http.server 8080
   ```
 
@@ -95,7 +95,7 @@ Click **Connect**. The dashboard polls `GET /adaptive_router/state` every
 In a second terminal, replacing `<your-master-key>` with the key printed in step 1:
 
 ```bash
-uv run python scripts/adaptive_router_demo/traffic.py \
+uv run python litellm/scripts/adaptive_router_demo/traffic.py \
     --proxy-url http://localhost:4000 \
     --api-key   <your-master-key> \
     --router    smart-cheap-router \

litellm/scripts/adaptive_router_demo/eval.py

@@ -9,7 +9,7 @@ For each test case:
   4. Prints PASS or FAIL with one line of reasoning.
 
 Run:
-  uv run python scripts/adaptive_router_demo/eval.py \
+  uv run python litellm/scripts/adaptive_router_demo/eval.py \
       --proxy-url   http://localhost:4000 \
       --api-key     "$LITELLM_MASTER_KEY" \
       --router      smart-cheap-router \

litellm/scripts/adaptive_router_demo/traffic.py

@@ -21,7 +21,7 @@ Why this shape:
     flow through routing + post-call hooks, so no API keys / no spend.
 
 Run:
-  uv run python scripts/adaptive_router_demo/traffic.py \\
+  uv run python litellm/scripts/adaptive_router_demo/traffic.py \\
       --proxy-url http://localhost:4000 \\
       --api-key   "$LITELLM_MASTER_KEY" \\
       --router    smart-cheap-router \\

litellm/scripts/benchmark_anthropic_messages_perf.py

@@ -13,11 +13,11 @@ streaming SSE wire format (``message_start`` -> ``content_block_delta`` ->
 can be compared without depending on real provider latency.
 
 Example:
-    uv run python scripts/benchmark_anthropic_messages_perf.py \
+    uv run python litellm/scripts/benchmark_anthropic_messages_perf.py \
         --label baseline --proxy-command ".venv/bin/litellm"
 
 Compare an already-running proxy:
-    uv run python scripts/benchmark_anthropic_messages_perf.py \
+    uv run python litellm/scripts/benchmark_anthropic_messages_perf.py \
         --no-start-proxy --label current
 """
 

litellm/scripts/benchmark_chat_completions_perf.py

@@ -6,11 +6,11 @@ from any checkout. That makes it useful for comparing tags/commits without
 depending on real provider latency.
 
 Example:
-    uv run python scripts/benchmark_chat_completions_perf.py \
+    uv run python litellm/scripts/benchmark_chat_completions_perf.py \
         --label current --requests 500 --concurrency 100
 
 Compare another checkout:
-    uv run python scripts/benchmark_chat_completions_perf.py \
+    uv run python litellm/scripts/benchmark_chat_completions_perf.py \
         --label v1.83.14-stable --litellm-dir /tmp/litellm-v1.83.14-stable
 """
 

litellm/scripts/benchmark_model_response_creator.py

@@ -6,8 +6,8 @@ isolate per-call cost. Driving the full wrapper adds threadpool logging,
 gc, and other noise that swamps microsecond-scale changes here.
 
 Example:
-    uv run python scripts/benchmark_model_response_creator.py --label baseline
-    uv run python scripts/benchmark_model_response_creator.py --label optimized
+    uv run python litellm/scripts/benchmark_model_response_creator.py --label baseline
+    uv run python litellm/scripts/benchmark_model_response_creator.py --label optimized
 """
 
 from __future__ import annotations

litellm/scripts/benchmark_ocr_callbacks.py

@@ -2,7 +2,7 @@
 """Measure serial sync/async OCR latency through a loopback HTTP provider
 
 Run each callback mode in a fresh process against an installed release wheel:
-python -I scripts/benchmark_ocr_callbacks.py --callbacks none --label before \
+python -I litellm/scripts/benchmark_ocr_callbacks.py --callbacks none --label before \
     --expected-transport rust --iterations 200 --warmup 20 --output before-none.json
 Repeat with --callbacks noop and with the candidate wheel in a separate venv
 """

litellm/scripts/benchmark_proxy_vs_provider.py

@@ -539,28 +539,28 @@ Examples:
   export PROVIDER_URL='https://api.openai.com/v1/chat/completions'
   export LITELLM_PROXY_API_KEY="$LITELLM_MASTER_KEY"
   export PROVIDER_API_KEY='sk-openai-key'
-  python scripts/benchmark_proxy_vs_provider.py
+  python litellm/scripts/benchmark_proxy_vs_provider.py
   
   # 2. Multiple runs for statistical accuracy (recommended)
-  python scripts/benchmark_proxy_vs_provider.py --runs 5
+  python litellm/scripts/benchmark_proxy_vs_provider.py --runs 5
   
   # 3. Realistic load testing with concurrency limit
-  python scripts/benchmark_proxy_vs_provider.py --max-concurrent 100 --requests 2000
+  python litellm/scripts/benchmark_proxy_vs_provider.py --max-concurrent 100 --requests 2000
   
   # 4. Quick test with fewer requests
-  python scripts/benchmark_proxy_vs_provider.py --requests 100
+  python litellm/scripts/benchmark_proxy_vs_provider.py --requests 100
   
   # 5. Parallel execution (not recommended - may affect accuracy)
-  python scripts/benchmark_proxy_vs_provider.py --parallel
+  python litellm/scripts/benchmark_proxy_vs_provider.py --parallel
   
   # 6. Custom timeout for slower endpoints
-  python scripts/benchmark_proxy_vs_provider.py --timeout 120
+  python litellm/scripts/benchmark_proxy_vs_provider.py --timeout 120
   
   # 7. Combined options for comprehensive testing
-  python scripts/benchmark_proxy_vs_provider.py --runs 3 --requests 500 --max-concurrent 50
+  python litellm/scripts/benchmark_proxy_vs_provider.py --runs 3 --requests 500 --max-concurrent 50
   
   # 8. Skip warmup (not recommended - may affect first request accuracy)
-  python scripts/benchmark_proxy_vs_provider.py --no-warmup
+  python litellm/scripts/benchmark_proxy_vs_provider.py --no-warmup
         """,
     )
     parser.add_argument(

litellm/scripts/benchmark_streaming_chunk_overhead.py

@@ -7,7 +7,7 @@ and Bedrock Converse (ModelResponseStream). A full proxy benchmark adds
 FastAPI, HTTP, and TCP latency, which dilutes the per-chunk CPU signal.
 
 Example:
-    uv run python scripts/benchmark_streaming_chunk_overhead.py \\
+    uv run python litellm/scripts/benchmark_streaming_chunk_overhead.py \\
         --streams 500 --chunks 200 --warmup 50 --repeats 5
 """
 

litellm/scripts/build_core_distribution.py

@@ -8,7 +8,7 @@ import tempfile
 from pathlib import Path
 from typing import Final
 
-ROOT: Final = Path(__file__).resolve().parents[1]
+ROOT: Final = Path(__file__).resolve().parents[2]
 SOURCES: Final = (
     "litellm",
     "litellm/rust",
@@ -57,7 +57,10 @@ def stage_core_distribution(source: Path, destination: Path, *, package: str = "
 
     def ignored_sources(directory: str, names: list[str]) -> set[str]:
         workspace_sources: Final = (
-            {"enterprise", "proxy-extras", "packaging", "rust", "cli", "ui", "vscode-extension", "deploy"}
+            {
+                "enterprise", "proxy-extras", "packaging", "rust", "cli", "scripts", "runtime",
+                "schema.prisma", "ui", "vscode-extension", "deploy",
+            }
             if Path(directory) == source / "litellm"
             else set()
         )

litellm/scripts/check_openapi_docs.py

@@ -1,7 +1,7 @@
 #!/usr/bin/env python3
 """List the POST operations in the proxy's OpenAPI document that Swagger UI cannot explain.
 
-Every finding is one line, ``<METHOD> <path> <code>``, so scripts/openapi_docs_gate.py can
+Every finding is one line, ``<METHOD> <path> <code>``, so litellm/scripts/openapi_docs_gate.py can
 count them per rule the same delta-vs-base way as the other lint gates. The rules:
 
 OA001  no request body schema (the handler reads ``request.body()`` and never told FastAPI

litellm/scripts/check_schema_compatibility.py

@@ -67,11 +67,18 @@ def main() -> int:
     options: Final = TypeAdapter(dict[str, str | None]).validate_python(vars(parser.parse_args()))
     base_commit: Final = options["base_commit"]
     base: Final = base_commit if base_commit else _merge_base(options["base_ref"] or "")
-    previous: Final = subprocess.run(
-        ("git", "show", f"{base}:schema.prisma"), capture_output=True, text=True, check=True
-    ).stdout
+    canonical: Final = subprocess.run(
+        ("git", "show", f"{base}:litellm/schema.prisma"), capture_output=True, text=True, check=False
+    )
+    previous: Final = (
+        canonical.stdout
+        if canonical.returncode == 0
+        else subprocess.run(
+            ("git", "show", f"{base}:schema.prisma"), capture_output=True, text=True, check=True
+        ).stdout
+    )
     changes: Final = check_breaking_changes(
-        parse_model_fields(previous), parse_model_fields(Path("schema.prisma").read_text())
+        parse_model_fields(previous), parse_model_fields(Path("litellm/schema.prisma").read_text())
     )
     if changes:
         sys.stderr.write("\n".join(changes) + "\n")

litellm/scripts/check_test_quality.py

@@ -1,10 +1,10 @@
 #!/usr/bin/env python3
 """Test-quality checker: the test-suite smells no linter enforces.
 
-Sibling of scripts/check_type_discipline.py, same output contract
+Sibling of litellm/scripts/check_type_discipline.py, same output contract
 (``path:line: CODE message``) and same stdlib-only constraint, aimed at the test
 tree instead of the package. Each rule is a shape the testing-strategy audit
-measured and named; scripts/test_quality_gate.py fails any change that grows the
+measured and named; litellm/scripts/test_quality_gate.py fails any change that grows the
 codebase total of one past its merge-base count.
 
 Rules

litellm/scripts/check_type_discipline.py

@@ -1172,6 +1172,8 @@ def collect_paths(raw: Iterable[str]) -> Iterator[Path]:
                         "litellm/packaging/",
                         "litellm/rust/",
                         "litellm/cli/",
+                "litellm/scripts/",
+                "litellm/runtime/",
                     )
                 )
             )
@@ -1186,6 +1188,8 @@ def collect_paths(raw: Iterable[str]) -> Iterator[Path]:
                 "litellm/packaging/",
                 "litellm/rust/",
                 "litellm/cli/",
+                "litellm/scripts/",
+                "litellm/runtime/",
             )
         ):
             yield p

litellm/scripts/ci_cd/cost_map_guard.py

@@ -2,7 +2,7 @@
 
 Every pull request whose diff against its merge base touches one of the two cost map files or the pins file gets the
 file checks: the files parse, the backup copy matches the root file, and every capability pinned in
-scripts/ci_cd/cost_map_pins.json (a value verified against a live provider call where the provider's model listing says
+litellm/scripts/ci_cd/cost_map_pins.json (a value verified against a live provider call where the provider's model listing says
 otherwise) still holds in the tree that merging the pull request into its base branch produces, so a branch cut before
 a pin was added cannot lower it, and a pull request that deletes the pins file fails too. A pull request that leaves
 all three untouched skips them, since merging it keeps the base branch's copies and its head tree only carries
@@ -27,7 +27,7 @@ from typing import Final
 
 COST_MAP_PATH: Final = "model_prices_and_context_window.json"
 BACKUP_PATH: Final = "litellm/model_prices_and_context_window_backup.json"
-PINS_PATH: Final = "scripts/ci_cd/cost_map_pins.json"
+PINS_PATH: Final = "litellm/scripts/ci_cd/cost_map_pins.json"
 PINS_PATHS: Final = (
     "litellm/scripts/ci_cd/cost_map_pins.json",
     "scripts/ci_cd/cost_map_pins.json",

litellm/scripts/ci_cd/run_migration.py

@@ -238,7 +238,7 @@ def create_migration(
         skip_freshness_check (bool): Skip the "branch is up to date" check.
             Only for intentional migrations against an older base.
     """
-    root_dir = Path(__file__).resolve().parents[2]
+    root_dir = Path(__file__).resolve().parents[3]
 
     if skip_freshness_check:
         print(
@@ -254,7 +254,7 @@ def create_migration(
         migrations_dir = (
             root_dir / "litellm" / "proxy-extras" / "litellm_proxy_extras" / "migrations"
         )
-        schema_path = root_dir / "schema.prisma"
+        schema_path = root_dir / "litellm/schema.prisma"
 
         # Create temporary PostgreSQL database
         with testing.postgresql.Postgresql() as postgresql:

litellm/scripts/db/backfill_daily_tool_spend.sql

@@ -26,7 +26,7 @@
 --    such buckets keep the writer's numbers and skip the backfill's.
 --
 -- Usage:
---   psql "$DATABASE_URL" -v cutover="'2026-07-25T00:00:00Z'" -f scripts/db/backfill_daily_tool_spend.sql
+--   psql "$DATABASE_URL" -v cutover="'2026-07-25T00:00:00Z'" -f litellm/scripts/db/backfill_daily_tool_spend.sql
 
 SET TIME ZONE 'UTC';
 

litellm/scripts/db/backfill_key_total_spend.sql

@@ -22,14 +22,14 @@
 --    budget_id links to a LiteLLM_BudgetTable row whose budget_duration IS
 --    NOT NULL (a linked budget resets the key's spend each period too). For
 --    those keys this script only recovers the current period;
---    scripts/db/backfill_key_total_spend_from_spend_logs.sql is an optional
+--    litellm/scripts/db/backfill_key_total_spend_from_spend_logs.sql is an optional
 --    follow-up that rebuilds the earlier periods from LiteLLM_SpendLogs.
 --
 -- 3. No proxy restart is needed. The proxy picks up the corrected values on
 --    its next read of each key.
 --
 -- Usage:
---   psql "$DATABASE_URL" -f scripts/db/backfill_key_total_spend.sql
+--   psql "$DATABASE_URL" -f litellm/scripts/db/backfill_key_total_spend.sql
 
 UPDATE "LiteLLM_VerificationToken"
 SET total_spend = spend

litellm/scripts/db/backfill_key_total_spend_from_spend_logs.sql

@@ -1,4 +1,4 @@
--- Optional follow-up to scripts/db/backfill_key_total_spend.sql. Run that
+-- Optional follow-up to litellm/scripts/db/backfill_key_total_spend.sql. Run that
 -- script first; this one rebuilds earlier budget periods for the keys it
 -- can only partially fix: keys whose spend resets each period, because their own
 -- budget_duration IS NOT NULL or because their budget_id links to a
@@ -38,7 +38,7 @@
 --    its next read of each key.
 --
 -- Usage:
---   psql "$DATABASE_URL" -f scripts/db/backfill_key_total_spend_from_spend_logs.sql
+--   psql "$DATABASE_URL" -f litellm/scripts/db/backfill_key_total_spend_from_spend_logs.sql
 
 -- Active keys whose spend resets (own budget_duration, or a linked
 -- LiteLLM_BudgetTable row with one). Rebuild from LiteLLM_SpendLogs,

litellm/scripts/db/partition_spend_logs.sql

@@ -32,7 +32,7 @@
 -- The spend-log cleanup job then verifies the table is partitioned and reclaims
 -- disk by dropping expired partitions instead of deleting rows. It also
 -- pre-creates upcoming partitions on each run. To roll back, see
--- scripts/db/unpartition_spend_logs.sql.
+-- litellm/scripts/db/unpartition_spend_logs.sql.
 
 BEGIN;
 

litellm/scripts/db/unpartition_spend_logs.sql

@@ -1,4 +1,4 @@
--- Rolls back scripts/db/partition_spend_logs.sql: converts the native
+-- Rolls back litellm/scripts/db/partition_spend_logs.sql: converts the native
 -- range-partitioned "LiteLLM_SpendLogs" table back into a plain,
 -- non-partitioned table matching the default LiteLLM schema.
 --

litellm/scripts/eval_compression.py

@@ -4,9 +4,9 @@ Prompt Compression Evaluation Harness
 Compare model performance on coding tasks with and without prompt compression.
 
 Usage:
-    python scripts/eval_compression.py --model gpt-4o --problems 5
-    python scripts/eval_compression.py --model claude-sonnet-4-20250514 --problems 12 --runs 3
-    python scripts/eval_compression.py --model gpt-4o-mini --padding-factor 50
+    python litellm/scripts/eval_compression.py --model gpt-4o --problems 5
+    python litellm/scripts/eval_compression.py --model claude-sonnet-4-20250514 --problems 12 --runs 3
+    python litellm/scripts/eval_compression.py --model gpt-4o-mini --padding-factor 50
 
 The harness runs each problem in two modes:
   1. **baseline** — raw prompt sent directly to the model.

litellm/scripts/gate_slot_lock.py

@@ -2,8 +2,8 @@
 """Machine-wide slot lock for this repo's heavy entrypoints.
 
 `make check`, `make lint`, and the standalone lint gates
-(scripts/ruff_strict_gate.py, scripts/type_discipline_gate.py,
-scripts/type_check_gate.py) each hold one of N machine-wide slots while they
+(litellm/scripts/ruff_strict_gate.py, litellm/scripts/type_discipline_gate.py,
+litellm/scripts/type_check_gate.py) each hold one of N machine-wide slots while they
 run, so however many sessions and worktrees share one machine, at most N of
 them execute a basedpyright/pytest/prettier storm at a time instead of all
 thrashing it at once. Slots are fcntl.flock files (macOS ships no flock(1)
@@ -22,7 +22,7 @@ itself. Any filesystem error fails open and the command runs unlocked: the
 lock is a courtesy to the machine, never a gate that may break a build (CI
 runs one job per machine, so there it only ever takes the instant path).
 
-CLI: python3 scripts/gate_slot_lock.py <command> [args...]
+CLI: python3 litellm/scripts/gate_slot_lock.py <command> [args...]
 """
 
 from __future__ import annotations

litellm/scripts/generate_provider_endpoints_support.py

@@ -11,7 +11,7 @@ from typing import Final
 
 from pydantic import JsonValue, TypeAdapter
 
-ROOT: Final = Path(__file__).resolve().parents[1]
+ROOT: Final = Path(__file__).resolve().parents[2]
 COST_MAP_PATH: Final = ROOT / "model_prices_and_context_window.json"
 SOURCE_PATH: Final = ROOT / "provider_endpoints_support.json"
 OUTPUT_PATHS: Final = (
@@ -218,7 +218,7 @@ def generate_files(
         sys.stderr.write(
             "provider endpoint support is stale in "
             + ", ".join(str(path) for path in stale_paths)
-            + "\nRun `uv run --no-sync python scripts/generate_provider_endpoints_support.py` to regenerate it\n"
+            + "\nRun `uv run --no-sync python litellm/scripts/generate_provider_endpoints_support.py` to regenerate it\n"
         )
         return 1
     for path in output_paths:

litellm/scripts/generate_trace_types.py

@@ -18,10 +18,10 @@ from typing import Final
 
 from pydantic import BaseModel, ConfigDict, JsonValue, TypeAdapter
 
-ROOT: Final = Path(__file__).resolve().parents[1]
-TOOLING: Final = ROOT / "scripts/trace_codegen"
+ROOT: Final = Path(__file__).resolve().parents[2]
+TOOLING: Final = ROOT / "litellm/scripts/trace_codegen"
 GENERATED: Final = ROOT / "litellm/tracing/generated"
-ASSETS: Final = ROOT / "scripts/lens_assets"
+ASSETS: Final = ROOT / "litellm/scripts/lens_assets"
 SCHEMAS: Final = TypeAdapter(dict[str, dict[str, JsonValue]])