fix: replace dead OpenRouter :free slugs with current valid ones + prune unkeyed providers

fix: working GitHub Models fallback + file limits that fit free-tier context

fix: point frontend to Render instead of HF Space

fix: batch cooldown all OpenRouter models on rate limit, suppress verbose errors

fix: add litellm to requirements.txt (missing deps)

fix: pin Python 3.12 for Render (crewai incompatible with 3.14)

restart: trigger HF Space rebuild

fix: suppress pydantic/litellm warnings, disable verbose tool logs

fix: resolve merge conflicts in main.py, remove bots.py

feat: add GitHub Models support, remove Groq dependency, add backend .gitignore

Deploy full Fitness AI Agents backend: auth, DB, watch sync, GPS routes, AI analysis

fix: serialize concurrent requests, pre-check cooldowns, prevent double file

Final updates

Final updates

fix: lower smart model max_tokens, filter litellm [RETRY] noise

fix: drop RAG search tools, fix rate limit detection

fix: add root-level Dockerfile required by HuggingFace Docker Spaces

fix: add HuggingFace Space metadata to README

chore: bust Docker layer cache to force full redeploy

fix: force caching=False in litellm call to stop cache_breakpoint re-injection

fix: monkey-patch litellm.completion to strip cache_breakpoint from messages

fix: always scan from index 0 so Groq is re-prioritized after cooldown

feat: module-level cooldown system for smart provider cycling

fix: drop unsupported params, rotate on BadRequestError

fix: disable litellm cache to stop Groq cache_breakpoint error

fix: allow tmp paths, reduce Groq TPM, suppress trace noise

fix: install litellm explicitly before crewai to prevent resolver override

fix: add litellm to requirements for groq/ and non-OpenAI provider support

fix 5 bugs: deprecated Groq models, retry-after parsing, 401 rotation, stale comment

remove Gemini from pools; Groq + OpenRouter is sufficient

add Groq + Gemini as primary free providers ahead of OpenRouter fallback

fix: 3 bugs — pipeline diagram, auto-retry guard, stop-during-retry

expand model pools: 4→8 fast, 3→7 smart with diverse providers

Quick style update

fix: promote gemma-3-27b-it to first SMART model; suppress CrewAI box-drawing noise

fix: set max_retries=0 on LLMs to prevent litellm double-retry on 429

fix: replace 404 model IDs with verified OpenRouter free-tier models

fix: filter SSE noise during rotation; amber [ROTATE]/[RETRY] log style; auto-retry once on full failure

fix: retry same model once on 429 before rotating; add api.md

Updated stuff

fix: replace nonexistent qwen3-30b-a3b; sleep retry_after+5s on 429 instead of 2s

fix: replace hermes-3 (no tool use) with tool-capable models; add 404 to rotation triggers

fix: replace Crucible-hosted models (deepseek-v4-flash, z-ai/glm) with NVIDIA/Qwen to avoid shared broken provider

Fix three bugs: redundant agent init, misleading rotate log, dead code

Rotate on 402/503/529 as well as 429

Add model rotation — auto-fallback on 429 across 4 fast / 3 smart models

Rebrand tab: title → DataFlow — AI Data Analysis, new favicon

Fix HF Space deployment and pipeline reliability

updates again

updates made