chore: refresh llms.txt and site metadata to current positioning - #72
Merged
Conversation
None of the hand-maintained metadata surfaces track docs content, so all of them still carried the pre-rewrite positioning. /api/docs/llms-full.txt picks up page bodies automatically, but its header does not — it comes from LLMS_PRODUCT_SUMMARY, as does /llms.txt. Positioning, everywhere: - LLMS_PRODUCT_SUMMARY, root metadata (title/description/openGraph/ twitter), the Organization + WebSite JSON-LD, and the OG image now say "context-aware LLM router that continuously improves your agent workflows" instead of "optimizes your agent for cost and performance". - Key Facts gain bitrouter/auto and the two improvement loops (online adequacy, offline `bitrouter optimize`). Factual corrections found while editing these files: - Local proxy port was http://localhost:8787 in llms.txt (3x), llms-shared.ts, and the MCP server's copy-paste snippet. The binary listens on 127.0.0.1:4356 per .cli-snapshot.json and quickstart. - llms.txt told agents to request `openai/gpt-4o` and `anthropic/claude-sonnet-4` in a copy-paste verify command. Neither id exists in .models-snapshot.json. - "200+ models" in the summary and OG image; the catalog holds 52, and bitrouter-vs-openrouter says "roughly 50 models, deliberately". Index coverage: added What is BitRouter?, Supported Models, vs LiteLLM, ACP Gateway, and a Usage section (CLI, TUI, Configuration, MCP Server, Agent Skills). All nine verified 200. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Answers "did the metadata update itself?" — no. Every hand-maintained surface still carried the pre-rewrite positioning, and three factual errors came along with it.
Only
/api/docs/llms-full.txtpicks up doc-page bodies automatically. Its header, and all of/llms.txt, come fromLLMS_PRODUCT_SUMMARY, which is hand-written.Positioning
LLMS_PRODUCT_SUMMARY, rootmetadata(title / description / openGraph / twitter), the Organization + WebSite JSON-LD, and the OG image all said "optimizes your agent for cost and performance with every run." They now say "context-aware LLM router that continuously improves your agent workflows" — matching the landing hero and #71.Key Facts gained
bitrouter/autoand the two improvement loops (online adequacy, offlinebitrouter optimize).Factual corrections found while editing these files
These were pre-existing and unrelated to the rewrite, but they live in the lines being changed and are wrong in the file whose entire job is being ingested correctly by agents:
http://localhost:8787http://127.0.0.1:4356llms-txt.ts×3,llms-shared.ts,lib/mcp/format.tsopenai/gpt-4o,anthropic/claude-sonnet-4anthropic/claude-opus-4.8,anthropic/claude-haiku-4.5llms-txt.tsllms-shared.ts,opengraph-image.tsxThe port is
4356per.cli-snapshot.json(×2) andquickstart.mdx(×6). Neither of the two model ids exists in.models-snapshot.json— and one of them was inside a copy-pastecurlthat llms.txt hands to agents as step 4, so an agent following the instructions got a bad request. The catalog holds 52 models;bitrouter-vs-openrouter.mdxsays "roughly 50 models, deliberately," so "200+" also contradicted our own comparison page.Index coverage
/llms.txtwas missing the page it now leads with. Added What is BitRouter?, Supported Models, vs LiteLLM, ACP Gateway, and a Usage section (CLI, TUI, Configuration, MCP Server, Agent Skills). All nine verified200against a local server.Not touched — needs your call
93.2% SWE-benchin the OG image. This number appears nowhere else in the repo and has no source I could find. It ships in every social preview. I left it rather than invent a replacement, but it should either get a citation or come out.<10ms Overhead(OG image) vs~5ms p50(llms-shared.ts). Not contradictory, but two different public numbers for the same thing.components/landing/compare/compare-programs.tsx:76also prints port8787. Left alone per "landing page is for illustration only."Depends on
bitrouter/autoreferences assume bitrouter/bitrouter#788 lands first, same as #71. No dependency on #65.Verification
vitest run— 83 passed, 11 filespnpm lint:docs— passes/llms.txtfetched from a local server: renders correctly, zero occurrences of8787,gpt-4o, or200+ models200🤖 Generated with Claude Code