Repository navigation
Conversation
|
hey @tmeire i think this fell off my radar because it's in draft status; did you still want our eyes on it? |
|
Hi @adamdotdevin you can definitly have a look at the PR and use it for inspiration. I'll see if I can rebase it, but the two-step process with the SDK generation into different repo's is annoying for something that is otherwise just a monorepo. I can work around for local dev with a go mod alias, but I would still have to split it into two PR's for what is one single logical unit. |
|
yeah, stainless isn't the most opensource friendly option for sdks but, fwiw, i'm happy to push updated sdks into your pr branch as needed (have done so for other prs) |
|
I'm going to close this PR. I have other things on my plate that are more urgent than reworking this PR. Feel free to use it as a starting point. |
… to the whitelist Review of 20e4d93 (HIGH 1, MEDIUM 3, LOW 5): - profile/AGENTS.md is loaded as instructions and still listed glm53-flash / deepseek-v4-flash-0731; it now lists the three served ids, and a test reads the list and compares it with the whitelist (red on 20e4d93's AGENTS.md) - a test pins the runtime model set of both copies to the whitelist, so a leftover model in one copy fails - README: the EVO-X2 router requires the key; the loopback default reaches no router today; no key export in the example (the key file is read automatically); the 262144 history is dated from 2026-09-13 (ai-cluster anomalyco#385); the block's source is ai-cluster anomalyco#614's evo1-router.clients.json - AGENTS.md says "no model-vendor egress" instead of "fully offline / air-gapped" (the router is reached over Tailscale) Agent-Lane: claude-code Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…red cor-local ids anywhere in AGENTS.md Review of 1847ae9 (Approve, LOW 4): - README recipe and the test comment no longer say offline / air-gapped: requests go over Tailscale to the EVO-X2 router - the AGENTS.md test also fails when any `cor-local/<id>` in the file names an id the lane does not serve (a fallback line with glm53-flash passed before; red now) - the 2026-09-13 split is credited to ai-cluster anomalyco#389 (requested in anomalyco#385); the long comment line is wrapped under 120 Agent-Lane: claude-code Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
No description provided.