diff --git a/docs/administration/deploy-digital-ocean.mdx b/docs/administration/deploy-digital-ocean.mdx index a72e08334..35eb2a20c 100644 --- a/docs/administration/deploy-digital-ocean.mdx +++ b/docs/administration/deploy-digital-ocean.mdx @@ -3,4 +3,6 @@ title: "Digital Ocean" description: "Deploy Arc to Digital Ocean" --- -Coming soon. + +This guide is coming soon. Deployment guides are currently in development. + diff --git a/docs/administration/deploy-fly-io.mdx b/docs/administration/deploy-fly-io.mdx index c8522faa5..118c562e6 100644 --- a/docs/administration/deploy-fly-io.mdx +++ b/docs/administration/deploy-fly-io.mdx @@ -3,4 +3,6 @@ title: "Fly.io" description: "Deploy Arc to Fly.io" --- -Coming soon. + +This guide is coming soon. Deployment guides are currently in development. + diff --git a/docs/administration/deploy-railway.mdx b/docs/administration/deploy-railway.mdx index 9b9c54f93..2afc35d6d 100644 --- a/docs/administration/deploy-railway.mdx +++ b/docs/administration/deploy-railway.mdx @@ -3,4 +3,6 @@ title: "Railway" description: "Deploy Arc to Railway" --- -Coming soon. + +This guide is coming soon. Deployment guides are currently in development. + diff --git a/docs/administration/deploy-render.mdx b/docs/administration/deploy-render.mdx index 45922c600..6528e1426 100644 --- a/docs/administration/deploy-render.mdx +++ b/docs/administration/deploy-render.mdx @@ -3,4 +3,6 @@ title: "Render" description: "Deploy Arc to Render" --- -Coming soon. + +This guide is coming soon. Deployment guides are currently in development. + diff --git a/docs/changelog/2026-03-07.mdx b/docs/changelog/2026-03-07.mdx index d55a56b74..97525bbb8 100644 --- a/docs/changelog/2026-03-07.mdx +++ b/docs/changelog/2026-03-07.mdx @@ -1,25 +1,33 @@ --- -title: "Models API, exe.dev sandboxes, and auto-install agent CLIs" +title: "Completions API, Sprites and exe.dev sandboxes, auto-install agent CLIs" date: "2026-03-07" --- -## exe.dev sandbox provider (beta) +## Completions API with SSE streaming -Arc can now run agent stages inside ephemeral [exe.dev](https://exe.dev) VMs as an alternative to Daytona sandboxes. The new `arc-exe` crate manages VM lifecycle through SSH — a management plane connection to `exe.dev` for creating and destroying VMs, and a data plane connection to the VM itself for command execution and file I/O. +Arc now exposes a `POST /completions` endpoint for single-turn LLM completions. You can use it for structured output via JSON Schema, one-off prompts, or building custom frontends on top of Arc's model routing. Streaming uses Anthropic-style SSE events (`message_start`, `content_block_delta`, `message_stop`, etc.), so you get tokens as they're generated rather than waiting for the full response. + +The CLI's `arc llm prompt` command can now target the server instead of calling providers directly: + +```bash +arc llm prompt "Summarize this file" --mode server +``` + +## Two new sandbox providers: exe.dev and Sprites + +Arc now supports three sandbox environments for running agent stages. In addition to Daytona, you can now use **exe.dev** VMs or **Sprites** (powered by Fly.io). + +exe.dev (beta) manages ephemeral VMs through SSH with separate management and data plane connections. Sprites uses the `sprite` CLI to provision Fly.io VMs with full sandbox lifecycle support. ```toml title="run.toml" [execution] -environment = "exe" +environment = "exe" # or "sprites" or "daytona" ``` ## Auto-install agent CLIs in sandboxes Agent CLI tools (like `claude` or `codex`) are now automatically detected and installed inside sandboxes at runtime when they're missing. This includes installing Node.js via NodeSource if needed. Previously, you had to build custom Dockerfiles that pre-installed every CLI tool your workflow used. -## Updated model catalog - -The default OpenAI model has been upgraded from gpt-5.3 to **gpt-5.4**. New models added to the catalog include **gpt-5.3-codex-spark** and **claude-sonnet-4-6**. Model defaults are now centralized in the catalog rather than hardcoded across crates. - **Verification API paths have changed.** `/verifications` is now split into `/verification/criteria` and `/verification/controls`. The run sub-resource moved from `/runs/{id}/verifications` to `/runs/{id}/verification`. @@ -28,9 +36,18 @@ To migrate: 2. Update any direct API calls to use the new paths + +**Removed API endpoints.** The following endpoints have been removed: +- `GET /projects` and `GET /projects/{name}/branches` — projects and branches are no longer exposed via the API +- `POST /workflows/{name}/runs` — use `POST /runs` instead (identical functionality) + +To migrate, regenerate your TypeScript client and update any direct API calls. + + ## More +- New `POST /completions` endpoint for single-turn LLM completions with SSE streaming and structured output via JSON Schema - New `GET /models` endpoint exposes the full LLM model catalog via the REST API with pagination - Verification API reorganized: `/verifications` split into `/verification/criteria` and `/verification/controls`; run sub-resource path is now `/runs/{id}/verification` @@ -38,13 +55,24 @@ To migrate: - Stage logs now stream in real-time while CLI agents are working, so you can see progress before a stage finishes - Sub-workflow context diffs no longer leak child engine-internal keys (`internal.*`, `graph.*`, `thread.*`) into the parent run +- Configurable git author identity for checkpoint commits via `[git.author]` in `server.toml` or `cli.toml` +- `arc llm prompt --mode server` routes prompts through the Arc server instead of calling providers directly - Turn and tool-call counts now display correctly in non-TTY mode + +- Default OpenAI model upgraded from gpt-5.3 to gpt-5.4 +- New models added to catalog: gpt-5.3-codex-spark and claude-sonnet-4-6 +- `apply_patch` parser now accepts bare `@@` hunk headers (compatibility with GPT-5.4 output) + + +- Fixed stall watchdog killing agents during long LLM reasoning turns — streaming events now correctly reset the timer +- Fixed OpenAI reasoning SSE events being silently swallowed, causing stall watchdog to fire during extended reasoning phases +- Fixed orphaned OpenAI reasoning items after context compaction causing "item provided without required following item" errors - Fixed push credentials for public repos in Daytona sandboxes - Fixed Daytona CLI installation with rootless Node.js install and better error messages