Feat/implement new and robust llm providers#146
Draft
patrickfleith wants to merge 54 commits into
Draft
Conversation
…okbook documentation
…publish to Hugging Face Hub
… to 100 samples, add life stage assignments, use multiple prompt variants, and push to Hugging Face Hub
…mization, reduced sample size, and simplified pipeline structure
- AGENTS.md/CLAUDE.md describe how agents work in this repo - add commit, decide, glossary, log, and write-manual skills
- PRD, roadmap, changelog, decisions, stack, and tasks under docs-agents/
- add reasoning profiles for Mistral (magistral) and Ollama thinking models, mapping reasoning_effort with an allowlist escape hatch where LiteLLM needs it - add per-provider capability resolvers and modality metadata (Gemini audio/video, Ollama image) - convert chat messages to Responses API input items and build data URIs for raw base64 media - fix rate-limit reservation, retry counting, and per-item batch retries; treat empty API keys as unset - cover reasoning resolution/forwarding and isolate provider tests from env side effects
- 11 runnable scripts plus README covering prompts, batching, structured output, metadata, multimodal, and rate limiting
- 11 runnable scripts plus README, including reasoning and multimodal image input
- 11 runnable scripts plus README, including magistral reasoning via reasoning_effort
- 11 runnable scripts plus README for local Ollama, with first-class reasoning on thinking-capable models - list the new openai/anthropic/mistral/ollama suites in the providers README
- ship sample_sunflower.jpg and send it as base64 so the request no longer depends on a routed provider fetching a remote URL
- record shipped capability-aware provider layer and example suites - add launch checklist: pipeline execution correctness and architecture - add provider hardening, tests, and launch documentation plan - capture feature expansion (audio/file/image io) and long-term items
- enable reasoning_effort and supports_reasoning on GEMINI_CHAT - forwarded natively for gemini/* (no allowlist); thinking=True maps to effort 'low'
- reliability: bounded retries, backoff growth, jitter range, timeout forwarding, RPM throttling, batch-retry ordering - gemini: capability resolution, reasoning_effort forwarding, thinking=True defaults to low effort
- 11 scripts mirroring the other providers, plus README and sample image - reasoning demos use native reasoning_effort; register gemini/ in providers index
- move reliability tests, gemini example suite, and first-class reasoning to Shipped - update task list: gemini manual-test task added, roadmap tasks marked done
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.