scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-09-07 08:26:10 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 36
46528 commits 4135 branches 1402 tags 2.1 GiB
Commit graph

1974 commits

Author SHA1 Message Date
mateo
6fa02887c4 feat(model_prices): add meta/muse-spark-1.3 and its contributor tier 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-02 22:05:17 +00:00
mateo-berri
f38a1ec129 Merge branch 'litellm_internal_staging' of https://github.com/BerriAI/litellm into litellm_azure_deepseek_v4_flash_0731 
# Conflicts:
#	litellm/model_prices_and_context_window_backup.json
#	model_prices_and_context_window.json
 2026-09-02 11:43:22 -07:00
Mateo Wang
4049a075bd
Merge pull request #39170 from BerriAI/litellm_registry_audit_2026_09_01 
fix(models): registry audit 2026-09-01: openai realtime and long-context tiers, mistral aliases, voyage, xai, fireworks, together, scaleway, azure ai, govcloud, azure gov, cloudflare whisper, deprecation dates
 2026-09-02 11:02:43 -07:00
mateo-berri
a9d3a0746c fix(models): price Azure DeepSeek V4 Flash 0731 from its own meters under the catalog id 2026-09-02 10:59:11 -07:00
mateo-berri
7a35c34303 fix(models): add the us-gov. geo inference profile keys for Claude Sonnet 5 and Opus 4.8 2026-09-02 10:30:14 -07:00
mateo
da23e0241d fix(models): add cloudflare whisper transcription pricing and pin govcloud pricing tests 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-02 15:28:53 +00:00
Yujong Lee
07cf9dc46f feat(models): add Azure DeepSeek V4 Flash 0731 2026-09-02 08:10:52 -07:00
mateo
b761277740 fix(models): drop inherited retirement dates from azure/us-gov entries pending a Government schedule source 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-02 15:09:49 +00:00
mateo-berri
6b83b16559 feat(gemini): day-0 pricing for gemini-3.8-flash 
Gemini 3.8 Flash launches today with the same promotional pricing, limits,
and thinking settings as Gemini 3.7 Flash, so the gemini/, vertex_ai/, and
bare cost map entries mirror the 3.7 Flash ones. Regression tests lock the
launch prices, the 4096-token cache minimum, and the gemini-3 thought
signature gate in for the new model.
 2026-09-02 08:04:14 -07:00
mateo
a7836ede15 fix(models): absorb open registry PRs: govcloud bedrock and mantle, azure gov, openai tiered long-context, scaleway, together qwen3.8, azure ai cache and kimi k2.7 code, azure mai deprecations 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-02 14:48:51 +00:00
mateo
23977fc290 Merge remote-tracking branch 'origin/litellm_internal_staging' into litellm_registry_audit_2026_09_01 2026-09-02 13:03:24 +00:00
James Liounis
e2c3f51c46
fix(search): forward search-tool params through the router, complete Parallel AI v1 param mapping (#37883) 
* fix(search): forward search-tool params through the router, complete Parallel AI v1 param mapping

SearchAPIRouter dropped every parameter configured on a search tool, forwarding
only per-request kwargs. Any tool-level setting (mode, max_results, ...) was
silently lost on the way to the adapter, for every search provider.

Also completes the Parallel AI v1 search surface: after_date, fetch_policy,
location and include_domains now nest under advanced_settings instead of being
sent as unknown top-level fields, responses preserve search_id / session_id /
warnings / raw excerpts, and search cost is derived from the request mode and
the provider's reported usage rather than a single flat rate.

* fix(parallel_ai): stop a caller from pricing its own search request

`_parallel_ai_usage` carries the provider's reported usage into cost
calculation. It was only written when the response contained a usage block, so
a caller could pass `_parallel_ai_usage=[{"name": "sku_search", "count": 0}]`
and, whenever the provider omitted usage, bill $0.00 instead of $0.005 — the
value also reached the upstream request body as an unknown field.

The key is now stripped from inbound params and written unconditionally from
the parsed response, so only the provider can populate it.

* fix(parallel_ai): price fast search mode correctly

* test(parallel_ai): fake search at HTTP boundary

* fix(parallel_ai): tolerate null search result fields

---------

Co-authored-by: khushishelat <shelatkhushi@gmail.com>
 2026-09-01 21:46:46 -07:00
mateo
d568bbe58d fix(bedrock): use tool fallback without forced tool_choice for claude-fable-5-1 structured output 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-01 20:19:46 +00:00
mateo
3e3e4d6970 fix(anthropic): use native structured output for claude-fable-5-1 on Vertex AI and Bedrock Invoke 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-01 20:01:03 +00:00
Devin AI
818fc5b913 Merge remote-tracking branch 'origin/litellm_internal_staging' into litellm_registry_audit_2026_09_01 2026-09-01 19:40:50 +00:00
Devin AI
24ee419c85 fix(models): registry audit 2026-09-01 for openai realtime, mistral aliases, voyage, xai, fireworks 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-01 19:28:04 +00:00
mateo
d6005a1876 merge: resolve conflict with litellm_internal_staging in anthropic transformation tests 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-01 19:23:47 +00:00
Mateo Wang
c50d83ece2
Merge pull request #39070 from BerriAI/litellm_bedrock_invoke_native_structured_output 
fix(bedrock): forward native structured outputs on Invoke instead of silently inlining the schema
 2026-09-01 12:19:15 -07:00
Mateo Wang
435433fa07
Merge pull request #39149 from BerriAI/litellm_qwencloud_provider_aliases 
feat(dashscope): add QwenCloud and Qwen AI Platform provider aliases
 2026-09-01 12:18:05 -07:00
Mateo Wang
4c3ef9ae0a
Merge pull request #39023 from BerriAI/litellm_add_azure_deepseek_v4_flash_0731 
feat: add Azure AI DeepSeek V4 Flash 0731 pricing
 2026-09-01 11:54:25 -07:00
mateo-berri
0042493bca Merge remote-tracking branch 'origin/litellm_internal_staging' into litellm_bedrock_invoke_native_structured_output 2026-09-01 11:50:05 -07:00
mateo
3c9ce458fd feat(anthropic): gate forced tool_choice for Fable 5.1 behind supports_forced_tool_use 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-01 18:46:03 +00:00
mateo-berri
f3792fb700 feat(dashscope): add qwencloud and qwen_ai_platform provider aliases 2026-09-01 11:20:36 -07:00
mateo
fb93db7791 feat(models): add Claude Fable 5.1 across Anthropic, Bedrock, Vertex AI, and Azure AI 
Adds claude-fable-5-1 cost map entries on the Anthropic API, Bedrock converse
(base, global, and us/eu geo inference profiles at the 10% regional premium),
Vertex AI, and Azure AI. Specs match Fable 5 (1M context, 128K output, $10/$50
per MTok, adaptive thinking always on, xhigh and max effort), except cache reads
land at $0.25 per MTok, a quarter of Fable 5's price and 0.025x base input
instead of the usual 0.1x.

Registers anthropic.claude-fable-5-1 in BEDROCK_CONVERSE_MODELS, lists the model
in the setup wizard, and extends the reasoning effort e2e grid. The partner cells
carry fail_reason markers until access on the CI accounts is confirmed.

Partner entries deliberately carry no deprecation_date: Anthropic publishes
retirement no sooner than 2027-09-01 for the first-party model, and the Foundry
and Vertex dates are not published yet.

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-01 18:07:06 +00:00
mateo-berri
06d4521fc0 fix(registry): add vertex veo 3.1 resolution tier pricing per vertex pricing page 2026-09-01 10:15:18 -07:00
Devin AI
9a1aebc146 fix(registry): declare databricks deepseek cache-write rate at the input rate per repo convention 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-01 13:50:21 +00:00
Devin AI
5263570e68 Add cerebras/zai-glm-4.7 deprecation_date per Cerebras deprecations page 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-01 13:33:31 +00:00
Devin AI
7bfa0d7fb4 Registry audit: Fireworks DeepSeek V4 Flash 0731 pricing, Databricks DeepSeek V4 entries, provider deprecation dates 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-09-01 13:15:22 +00:00
Devin AI
40738355b4 Merge remote-tracking branch 'origin/litellm_internal_staging' into devin_ai_1788201394-veo31-pricing-tiers 2026-09-01 13:03:19 +00:00
Tin Chi Lo
a27e12367e fix(bedrock): forward native structured outputs on Invoke instead of silently inlining the schema 2026-08-31 23:19:22 -07:00
Mateo Wang
7a02e4163f
Merge pull request #38913 from BerriAI/litellm_gigachat_passthrough_25886 
feat(gigachat): add native API passthrough routes with spend logging
 2026-08-31 15:47:27 -07:00
mateo-berri
eb00986f18 Merge remote-tracking branch 'origin/litellm_internal_staging' into litellm_gigachat_passthrough_25886 
# Conflicts:
#	osv-scanner.toml
 2026-08-31 15:25:10 -07:00
Mateo Wang
ac206518a0
Merge pull request #38881 from Lee-Si-Yoon/friendli/glm-5.3 
feat(friendli): add zai-org/GLM-5.3 model pricing
 2026-08-31 15:18:45 -07:00
Mateo Wang
336269cfec
Merge pull request #38597 from BerriAI/litellm_fix_nova_sonic_realtime_user_asr_usage 
fix(bedrock): surface Nova Sonic user transcripts, speech events, and usage in realtime API
 2026-08-31 15:11:31 -07:00
mateo-berri
abbccd3fd6 Merge branch 'litellm_internal_staging' of https://github.com/BerriAI/litellm into friendli/glm-5.3 
# Conflicts:
#	litellm/model_prices_and_context_window_backup.json
#	model_prices_and_context_window.json
 2026-08-31 15:11:06 -07:00
Mateo Wang
b3a1dd1115
Merge pull request #38880 from Lee-Si-Yoon/friendli/glm-5.3-flash-v2 
feat(friendli): add zai-org/GLM-5.3-Flash model pricing
 2026-08-31 15:07:06 -07:00
Yujong Lee
be5997f366 feat: add Azure AI DeepSeek V4 Flash 0731 pricing 2026-08-31 14:49:43 -07:00
Devin AI
6809d537f0 merge: litellm_internal_staging into litellm_fix_nova_sonic_realtime_user_asr_usage 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-08-31 21:42:06 +00:00
mateo-berri
bd794f9f18 fix(friendli): track GLM-5.3 discounted live pricing and declare effort levels 2026-08-31 13:26:37 -07:00
mateo-berri
a90fb538bf fix(friendli): declare GLM-5.3-Flash reasoning efforts as explicit levels 2026-08-31 13:25:51 -07:00
mateo-berri
59732f068b Merge remote-tracking branch 'origin/litellm_internal_staging' into litellm_gigachat_passthrough_25886 2026-08-31 13:16:41 -07:00
Devin AI
c344c7a66b fix(registry): add zai/glm-5.2, together Qwen3.8-Flash, cerebras/gemma-4-31b, elevenlabs/scribe_v2 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-08-31 19:11:04 +00:00
Devin AI
7b4b92f54f fix(registry): update veo 3.1 pricing with resolution tiers 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-08-31 18:37:26 +00:00
mateo-berri
8a6f47a6d4 merge: litellm_internal_staging into litellm_fix_nova_sonic_realtime_user_asr_usage 2026-08-31 10:17:12 -07:00
Devin AI
4291afbfa5 fix(registry): correct OpenAI preview shutdown dates, add whisper/transcribe and Bedrock/Vertex deprecation dates 
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
 2026-08-31 13:08:06 +00:00
Devin AI
f111262e54 Merge remote-tracking branch 'origin/litellm_internal_staging' into devin_ai_1788036971-stale-cost-map-sources 2026-08-31 13:03:35 +00:00
siyoon
e7bfe99cd3 feat(friendli): add zai-org/GLM-5.3 model pricing 
Per https://api.friendli.ai/serverless/v1/models:
- $1.40 input / $4.40 output / $0.26 cached input per MTok
- 1M context, 1M max output, reasoning with effort low/high/max
  (per HF chat_template.jinja: low/high honored, anything else -> max)
- tool calling, parallel tool calls, structured output, prompt caching
- text-only (no vision), flagship GLM model
 2026-08-30 14:23:36 +09:00
siyoon
3ea4b715ba feat(friendli): add zai-org/GLM-5.3-Flash model pricing 
Per https://api.friendli.ai/serverless/v1/models:
- $0.15 input / $0.50 output / $0.03 cached input per MTok
- 1M context, 1M max output, reasoning with effort low/high/max
  (per HF chat_template.jinja: low/high honored, anything else -> max)
- tool calling, parallel tool calls, structured output, prompt caching
- image + video input (native multimodal)
 2026-08-30 14:22:41 +09:00
mateo-berri
70e2f4e68f Merge remote-tracking branch 'origin/litellm_internal_staging' into litellm_gigachat_passthrough_25886 
# Conflicts:
#	litellm/llms/gigachat/chat/transformation.py
 2026-08-29 22:08:54 -07:00
Mateo Wang
42d8360f29
Merge pull request #38820 from BerriAI/litellm_fix_together_sync_output_ceiling 
fix(together_ai): stop writing context_length as max_output_tokens in the serverless sync
 2026-08-29 16:44:56 -07:00