#general
Hi team — I think I found a prompt-caching bug in the backend and couldn't open a GitHub issue (interactions are restricted to prior contributors), so posting here.
TL;DR: On the Bedrock path, applyPromptCacheBreakpoints attaches ttl:"1h" to cache breakpoints for Claude 4.6 models, but on Bedrock only the 4.5 generation supports the 1-hour TTL — 4.6 is 5-minute only. Bedrock rejects the request (it doesn't silently downgrade), so a Bedrock + Opus/Sonnet 4.6 chat fails.
Where: platform/backend/src/routes/chat/normalization/apply-prompt-cache.ts:24
// comment says "4.5 and newer support a 1-hour cache TTL"
const ONEHOURCACHE_MODEL = /claude-(?:sonnet|haiku|opus)-4-[5-9](?!\d)/;
This matches 4.5–4.9, so 4.6 gets tagged 1h-capable. useOneHour (lines 90-94) then writes ttl:"1h" into the Bedrock cachePoint regardless of provider.
Why it's wrong (Bedrock-specific): AWS's docs list the 1h TTL set by exact version — only Opus/Sonnet/Haiku 4.5; the 4.6 rows say 5-minute only. (Note: on the Anthropic-direct API 4.6 does support 1h, so the bug is specific to the Bedrock path.) Unsupported TTL is a hard error on Bedrock — see the parallel report openclaw/openclaw#21986.
Repro: your own unit test already encodes it — apply-prompt-cache.test.ts:282 uses model: "us.anthropic.claude-opus-4-6-v1:0" and asserts ttl:"1h".
Fix: narrow the 1h allowlist (Bedrock → 4.5 generation only) and default to the 5-minute TTL for anything not on it, so newer/unknown models degrade to 5m instead of failing. Happy to open a PR with the patch + updated test if you can whitelist my account for contributions — just let me know.
(Found while doing academic research on prompt-caching bugs in OSS LLM apps; full writeup available if useful.)
Quick question, has anyone tried routing ADK agents through Archestra's LLM Proxy ? I am able to talk to the agent but images/files are not going through, it seems like they are dropped and the agent only receives text. Anyone has any experience with routing different clients through the LLM proxy ?
I am using LiteLLM in order to route through Archestra since ADK does not seem to provide a way to change the
api_base. But I might be completely wrong in my approach, any help is appreciated.from google.adk.agents import Agent
from google.adk.models.lite_llm import LiteLlm
ARCHESTRA_LLM_PROXY_ID = "bc0ecdc9-ff33-43a1-a703-73a1fd1c80fb"
ARCHESTRA_MODEL_ROUTER_URL = f"<http://localhost:3000/v1/model-router/{ARCHESTRA_LLM_PROXY_ID}>"
root_agent = Agent(
name="test_agent",
model=LiteLlm(
model="gemini-2.5-flash",
api_base=ARCHESTRA_MODEL_ROUTER_URL,
api_key=os.getenv("ARCHESTRA_API_KEY"),
),
description=(
"Agent to talk to."
),
instruction=(
"You are a helpful agent who can have a discussion with the user."
),
)```👀1
15 replies
Hey everyone!
I'm Hamza Nazir, a MERN Stack Developer focused on building full-stack web applications with React, Node.js, Express, and MongoDB.
I joined Archestra to learn, connect with other developers, and contribute to open-source projects. Looking forward to collaborating and building with the community!
:archestra-love:1
2 replies
Hello Everyone! My name is Mike, I’m a fullstack dev. Most of my professional work has been with third-party development companies, so I thought this would be a great opportunity for me to help out the community! My primary skills are with PERN/MERN stacks, as well as Flask backends and Unity development, with React being my strongest skill. I would love to start contributing and am still familiarizing myself with the Archestra platform. If anyone has something important that needs a set of eyes on it, please do not hesitate to reach out!
👋2
What’s new in Archestra: from v1.2.62 to v1.2.71 🚀
We shipped quite a lot over the past 2 weeks, so here’s a roundup of the most notable changes.
:speech_balloon: Chat & agents
More models to choose from, more control over what the agent does each turn, and a clearer view of what it’s actually doing.
• GitHub Copilot is now a supported LLM provider: route Copilot models through Archestra with the usual virtual keys, cost limits and policies (#5539)
• Agent lifecycle hooks: run your own scripts at chat lifecycle events (SessionStart / PreToolUse / PostToolUse) inside the conversation sandbox, Claude-Code style (#5424)
• Context window visualizer: click the context indicator in chat to see exactly what’s filling the model’s context, with an estimated per-turn cost (#5260)
• MCP elicitation pass-through: gateway tools can now prompt the caller for input mid-call (#5494)
🔧 Connectors & messaging
Pull more of your sources into Archestra, and let agents work where your team already talks.
• New Perforce (Helix Core) knowledge connector (#5503)
• MS Teams replies to any message in a thread it was mentioned in, plus file/image upload in 1:1 chats (#5470)
• Slack sticky-thread auto-reply (#5490)
• Rebuilt client connection setup: step-by-step wizard, one-command setup, Windows PowerShell scripts (the new connection page itself is in beta) (#5516)
🔓 Teams, security & enterprise auth
Things that start matter once Archestra is more than one person: who can do what, what the agent’s sandbox can reach, and how clients prove who they are.
• Team roles: first-class admin/member roles per team, reworked team management, cleaner invite signup (#5425)
• Tighter network controls: per-environment sandbox egress, plus egress policy enforced on remote MCP servers (#5682)
• Enterprise OAuth: client-credentials and authorization_code grants for MCP gateways and LLM OAuth clients (#5632)
:archestra-love: In beta:
• MCP Apps: build, run and share your own apps right in chat, ARCHESTRAAPPSENABLED (#5437)
• Projects + My Files: group chats into projects, with a persistent file system on the sandbox, ARCHESTRAPROJECTSENABLED (#5599)
Read-only live mirror of Archestra.AI Slack
👋Join the discussion withAI enthusiasts!