enhance system prompts with updated guidelines and structured capabilities for user interactions

enhance system prompts with detailed guidelines and direct hyperlinks for user actions

enhance system prompts with direct hyperlinks for user actions

enhance prompts for better clarity and user guidance

Refactor OCR service to implement a resilient multi-provider router

fix dashboard isues

Deploy Full Engine

add open router models

enhance dash styles

F:\Aoun Grad Project\AwnAiService\AwnDeploy

enhance rag

enhance system prompt

fix routing isue

fix error

fix chat error

disable non work models

Allow OpenRouter free models with BYOK key, remove paid-only filtering and forced fallback block.

fix isue

feat: Hardened routing layer with Redis-backed rate limiting, distributed circuit breakers, bulkheads, and full OTel/Prom observability with Grafana alerts

feat: implement v4 resilient multi-provider LLM routing with smart scoring, circuit breakers, self-healing, dynamic models, and rate limiting

fix(llm): resolve production cascading failure and harden routing architecture

fix(llm): resolve NameError in gemini_client, update provider model lists, and fix 503 error propagation

refactor(llm): implement resilient multi-provider router, fix missing settings imports, and add OpenRouter fallback

refactor(llm): implement resilient multi-provider router, add circuit breaker, update models, and enforce 503 HTTP failovers

refactor(ai): remove all dead models and HF fallback to pure functional state

feat(ai): massively expand Groq and Gemini model fallback pool with separate quotas to eliminate rate limit exhaustions

fix(ai): use guaranteed free serverless models for HF and update Gemini aliases

fix(ai): update HF base URL to the new router architecture

fix(ai): Remove dead Groq/Gemini models and properly deploy OpenAI + HuggingFace multi-fallback clients

feat(ai): unify chat fallback logic with OCR architecture by adding OpenAI and HuggingFace providers

fix(ai): Implement LLM Router with Groq model pool & Gemini Fallback to protect against API rate limits and model deprecations

feat: implement multi-model LLM router with Groq and Gemini failover for rate limit protection

fix: disable automatic retries in Groq SDK to resolve 30s latency blocking

fix: pass family_id and access_token to _execute_stream to fix NameError

fix: inject credentials into tool calls during streaming to fix NoneType error

fix: polite fallback messages and fix Voice API 500 unpack error

fix: tightly forbid AI from exposing raw function XML tags to user

fix: handle null arguments from Groq when tools have no properties

fix: remove internal tokens from AI tools schema to fix validation 400

fix: add logging for fallback errors

fix: AI hallucination and add get_detailed_requests tool

feat: pass user access_token to .NET API for native authorization

feat: complete prompt rewrite and FAQ update for agent capabilities

fix: filter garbage history entries + anti-hallucination rules for tools

fix: add tool awareness to system prompt - LLM now uses agent tools instead of rejecting

feat: clean deployment with bug fixes and stability improvements