ent-v2.2.3-base
maximhq/bifrostent-v2.2.3-baseSep 24, 2026by akshaydeo
Upgrading?
AI Summary
Enterprise base update containing numerous fixes for routing, session affinity, OpenAI/GPT-6 handling, and telemetry.
Key Highlights
- Cherrypicked key selection to fallbacks and routing key pins with routes
- Session affinity fixes including route matching and key pin movement
- GPT-6 support and OpenAI reasoning/params fixes
- Telemetry user labels toggle fix
New Features
- Cherrypicked key id selection to fallbacks
- Move routing key pins with their routes on session affinity
- Carry route-level key pins through session affinity reorder
- Drop session bindings on request failure
- Add session affinity matrix tests
- Update UI path for user labels toggle
- Merge multi-part system prompts for Gemini-backed endpoints
- OpenAI reasoning and params fixes
- OpenAI prompt cache breakpoint fix
- Responses type fixes
- Async tool execution in responses
- GPT-6 sol and luna support
- Harness tests for GPT-6
- OpenRouter cache breakpoint fix
- Bedrock encrypted content decoding fix
- Kimi + codex fixes
- Tool_search support for gpt-6 models
Full Release Notes
## What's Changed * cherrypicked key id selection to fallbacks by @akshaydeo in https://github.com/maximhq/bifrost/pull/7470 * fix: move routing key pins with their routes when session affinity reorders the chain so a pinned key is never looked up under the wrong provider by @Pratham-Mishra04 in https://github.com/maximhq/bifrost/pull/7468 * fix: carry route-level key pins through session affinity reorder and add e2e session-affinity matrix tests by @Pratham-Mishra04 in https://github.com/maximhq/bifrost/pull/7469 * fix: drop session bindings the request followed into a failure and match route by provider and model together by @Pratham-Mishra04 in https://github.com/maximhq/bifrost/pull/7473 * test: add session affinity matrix cases 12–13 (binding names provider+model, binding forgotten after key-selection failure) and new harness routing rules (`anthropic-first`, `anthropic-first-other-model`, `vertex-badkey`) by @Pratham-Mishra04 in https://github.com/maximhq/bifrost/pull/7474 * fix: update UI path to correctly handle user labels toggle by @roroghost17 in https://github.com/maximhq/bifrost/pull/7490 * fix(databricks): merge multi-part system prompts for Gemini-backed endpoints by @impoiler in https://github.com/maximhq/bifrost/pull/7461 * fix: openai reasoning and params fixes by @TejasGhatte in https://github.com/maximhq/bifrost/pull/7239 * fix: openai prompt cache breakpoint by @TejasGhatte in https://github.com/maximhq/bifrost/pull/7240 * fix: responses type fixes by @TejasGhatte in https://github.com/maximhq/bifrost/pull/7241 * feat: async tool execution in responsses by @TejasGhatte in https://github.com/maximhq/bifrost/pull/7242 * fix: gpt-6 sol and luna support none effort by @TejasGhatte in https://github.com/maximhq/bifrost/pull/7492 * tests: harness tests gpt-6 by @TejasGhatte in https://github.com/maximhq/bifrost/pull/7496 * fix: openrouter cache breakpoint and harness fix by @TejasGhatte in https://github.com/maximhq/bifrost/pull/7521 * 2.2.3 changelog by @akshaydeo in https://github.com/maximhq/bifrost/pull/7528 * bedrock encrypted content decodding fix by @akshaydeo in https://github.com/maximhq/bifrost/pull/7529 * kimi + codex fix. fixes mcp tool name length by @akshaydeo in https://github.com/maximhq/bifrost/pull/7530 * tool_search support for gpt-6 models for non-openai providers by @akshaydeo in https://github.com/maximhq/bifrost/pull/7534 **Full Changelog**: https://github.com/maximhq/bifrost/compare/ent-v2.2.2-base...ent-v2.2.3-base
Related Repositories
ollama/ollama★ 184,214
Get up and running with Kimi, GLM, MiniMax, DeepSeek, gpt-oss, Qwen, Gemma and other models.
fatedier/frp★ 110,053
A fast reverse proxy to help you expose a local server behind a NAT or firewall to the internet.
infiniflow/ragflow★ 92,614
RAGFlow is a leading open-source Retrieval-Augmented Generation (RAG) engine that fuses cutting-edge RAG with Agent capabilities to create a superior context layer for LLMs
gohugoio/hugo★ 90,198
The world’s fastest framework for building websites.
junegunn/fzf★ 83,716
:cherry_blossom: A command-line fuzzy finder