What's Changed
- Map reasoning tokens across OpenAI-compatible extractors and enforce coverage by @dsfaccini in #580
- Fix four broken pipeline paths and close the gaps that hid them by @dsfaccini in #536
- Update OpenAI GPT-5.5 and GPT-5.6 prices by @Kludex in #592
- Keep Claude Sonnet 5 launch prices and update Voxtral Small by @Kludex in #593
- Split Mistral model generations and correct their prices by @Kludex in #595
- Fix order-dependent price lookup: cache write key must match its read key by @Hrafz in #560
- Extract Azure Realtime cached audio usage by @Kludex in #597
- Correct context windows for
claude-opus-4-6andclaude-sonnet-4-5by @xyzs996 in #585 - Include Gemini 2.5 Flash context window by @Kludex in #603
- Document aggregate cached input usage by @Kludex in #602
- Restrict Command A pricing to known models by @Kludex in #604
- Replace yargs with Node argument parsing by @Kludex in #605
- Add Mistral OCR pricing by @Kludex in #606
- Cache provider resolution in data snapshots by @Kludex in #599
- Support mapping usage inputs by @Kludex in #608
- Match Pixtral and Mixtral models to Mistral by @Kludex in #609
- Refresh Avian model pricing by @Kludex in #610
- Expand agentic price checks to direct providers by @Kludex in #611
- Restore 100% Python coverage by @dsfaccini in #598
- Add Gemini 2.5 TTS pricing by @Kludex in #612
- Apply Groq transcription billing minimums by @Kludex in #589
- Add OpenAI and Mistral transcription prices by @Kludex in #590
- Consolidate DeepSeek V4, Grok, and GPT-5.5 pricing updates by @dsfaccini in #615
- Use one JavaScript constraint shape by @Kludex in #613
- Add Zhipu GLM-5.3 and GLM-5.3 Flash pricing by @dsfaccini in #617
- Match compact dated model references by @dsfaccini in #616
- Revert JavaScript constraint shape changes by @dsfaccini in #618
- Exclude generated JSON from Macroscope review by @strawgate in #621
- feat: staged triage pilot on the pydantic-ai gh-aw engine by @dsfaccini in #620
- Match Azure AI Foundry host for Anthropic models by @Kludex in #408
- Derive the release version from the git tag by @Kludex in #622
- Add cache_write_tokens mapping to Azure's responses/chat extractors by @kecbigmt in #596
- Update Gemma 4 31B pricing for Google and OpenRouter by @Kludex in #623
- Add OpenRouter long-context tiers, dated ids and GPT-5.6 entries for OpenAI models by @Kludex in #625
- Add Moonshot Kimi K2.7 Code HighSpeed pricing by @Kludex in #626
- Add Cohere Aya Expanse and Command Light pricing by @ayaangazali in #472
- Refresh Cerebras provider: add Gemma 4 31B and GLM 4.7, deprecate delisted models by @ayaangazali in #469
- Add QuickSilver Pro provider by @raullenchai in #557
- Point stale pricing URLs at pages that expose prices to plain HTTP by @Kludex in #629
- Add web search tool call pricing to OpenAI models by @Kludex in #630
- Price Azure AI Foundry marketplace models through their origin providers by @Kludex in #633
- Add file search tool call pricing to OpenAI models by @Kludex in #631
- Correct the original gpt-4o snapshot and Devstral Small prices by @Kludex in #636
- Add Cloudflare Workers AI pricing by @Kludex in #637
- Add provider integration skill by @Kludex in #638
- Infer Cloudflare native model from request URL by @Kludex in #639
- Add Doubleword models: Kimi K3, GLM 5.2, Qwen3.8 27B, DeepSeek V4 Flash 0731 by @kylesf in #624
- Exclude generated bundled package data from Macroscope review by @strawgate in #640
- Share model metadata across provider offerings by @adtyavrdhn in #465
- Fill
openrouterrecord context windows from the OpenRouter API by @adtyavrdhn in #641 - Add Cursor model pricing by @Kludex in #642
- Price xAI Realtime voice usage by @Kludex in #600
- Add Arcee provider pricing by @Kludex in #644
- Fill novita record context windows from the Novita API by @adtyavrdhn in #643
- Fill mistral record context windows from the Mistral API by @adtyavrdhn in #648
- Fill groq record context windows from the Groq API by @adtyavrdhn in #649
- Fill openai record context windows from OpenAI's model docs by @adtyavrdhn in #650
- Fill aws record context windows from the Bedrock model cards by @adtyavrdhn in #651
- Fill
azurerecord context windows from Microsoft's model docs by @adtyavrdhn in #652 - Fill remaining single-record context windows (cerebras, moonshotai, minimax) by @adtyavrdhn in #656
- Fill
perplexityrecord context windows from Perplexity's model cards by @adtyavrdhn in #655 - Fill
x-airecord context windows from xAI's model docs by @adtyavrdhn in #654 - Fill
deepseekrecord context windows from DeepSeek's API docs by @adtyavrdhn in #653 - Expose a default api_flavor on generated HuggingFace and OVHcloud providers by @Kludex in #634
- Refuse to overwrite provider YAML from an empty or sharply shrunken importer response by @Kludex in #635
- Add Baseten Model API pricing by @Kludex in #660
- Add native Go pricing package by @Kludex in #658
New Contributors
- @Hrafz made their first contribution in #560
- @xyzs996 made their first contribution in #585
- @kecbigmt made their first contribution in #596
- @ayaangazali made their first contribution in #472
- @raullenchai made their first contribution in #557
- @kylesf made their first contribution in #624
Full Changelog: v0.1.4...v0.1.5