Odel
nyuchi docs

nyuchi docs

@nyuchiMDXUpdated 1w ago

Search, read, and ask the Nyuchi docs (docs.nyuchi.com); send feedback or raise issues.

View on GitHubSign in to connect
Server endpointStreamable HTTPNo authProbed
https://docs.nyuchi.com/mcp

This is the third-party server itself — Odel doesn't run it. Hitting this URL directly talks straight to the upstream server with no auth or proxying. Connect through Odel to front it with managed auth.

nyuchi-docs

Nyuchi engineering documentation — how things are done at Nyuchi, and how to use the Mzizi tools from a Nyuchi project. Published at docs.nyuchi.com.

This repo is a pnpm workspace with these packages:

PackagePathWhat it does
sitesite/The Astro + Starlight docs site itself. Ships as a Cloudflare Worker with Static Assets.
@nyuchi/nyuchi-docs-searchnyuchi-docs-search/Publishable npm package: cmdk-style search modal + Ask-AI tab for Starlight sites.
shamwari-docs-aishamwari-docs-ai/Cloudflare Worker — the Ask-AI chat proxy (SSE).
nyuchi-docs-mcp-workernyuchi-docs-mcp-worker/Cloudflare Worker nyuchi-docs-mcp — the docs MCP server at docs.nyuchi.com/mcp.

Companion site

bundu-labs/bundu-docs covers the Bundu Foundation's outward-facing projects — the Mzizi product, the Ubuntu doctrine, and the Bundu brand system. It installs @nyuchi/nyuchi-docs-search from npm and points at the nyuchi-docs-mcp worker.

Sections (site/src/content/docs/)

  • platform/ — the product guide for the Nyuchi platform.
  • api/ — API Docs: the /v1 gateway, WorkOS authentication, console-managed API keys, security, and the product namespaces.
  • analytics/ — dashboards, reports, and connecting data sources.
  • kweli/ — Mukoko Kweli product guides: verification, cross-app how-to, open data, data quality, design system.
  • mukoko-weather/ — Mukoko Weather user guide and stations.
  • integrations/ — connectors, webhooks, the docs MCP server, and the Mukoko Events MCP server.
  • identity/ — WorkOS, accounts.mukoko.com (the AuthKit issuer), SSO, JWTs.
  • console/ — the Nyuchi Console at platform.nyuchi.com.
  • tools/ — the cross-repo tools directory: every skill, CLI, and MCP server across the Nyuchi and Bundu repos.
  • mzizi-tools/ — mzizi-mcp, mzizi-cli, mzizi-skills, the DNA double-helix architecture, registry health, and the A2A design.
  • deployment/ — Cloudflare, Vercel, and Supabase deployment patterns.
  • conventions/ — PR doctrine, commit doctrine, repo-naming rules.

Develop

pnpm install
pnpm dev            # site only
pnpm -r build       # all packages
pnpm -r test        # all packages

Site dev server: http://localhost:4321. Content lives in site/src/content/docs/; the sidebar is configured in site/astro.config.mjs.

Search + Ask AI

The search modal opens with ⌘K / Ctrl+K. The Ask AI tab streams answers from shamwari-docs-ai (Cloudflare Worker) with retrieval-grounded citations. To enable it locally, copy site/.env.example to site/.env:

cp site/.env.example site/.env
# .env:
# PUBLIC_SHAMWARI_AI_URL=https://shamwari-docs-ai.nyuchi.workers.dev

Deploy

Both workers in this repo deploy via Cloudflare Workers Builds — the Cloudflare GitHub App is connected to nyuchi/nyuchi-docs with one trigger per worker (root directory points at the worker package). No GitHub Actions deploy workflow, no CLOUDFLARE_API_TOKEN repo secret.

  • nyuchi-docs (site) — root site/, ships as a Cloudflare Worker with Workers Static Assets. Live at https://nyuchi-docs.nyuchi.workers.dev. Custom domain docs.nyuchi.com is attached via the Workers custom-domain API in a separate cutover step (apex still points at the legacy Mintlify-on-Vercel deployment until then).
  • shamwari-docs-ai — root shamwari-docs-ai/, thin proxy in front of Cloudflare AI Search. Live at https://shamwari-docs-ai.nyuchi.workers.dev. See shamwari-docs-ai/README.md for the per-corpus AI Search instance setup (managed via REST API).
  • nyuchi-docs-mcp — root nyuchi-docs-mcp-worker/, the docs MCP server. Live at https://nyuchi-docs-mcp.nyuchi.workers.dev and routed from docs.nyuchi.com/mcp*. Needs its own Workers Builds trigger (root directory nyuchi-docs-mcp-worker/).

Well-known and machine-readable endpoints

docs.nyuchi.com serves these outside the docs tree. Most are static files in site/public/; security.txt is generated per request by the site worker.

PathServed fromWhat it is
/robots.txtsite/public/robots.txtCrawl policy + sitemap pointer. Everything here is meant to be indexed.
/llms.txtsite/public/llms.txtMachine-readable site index for LLMs.
/AUTH.mdsite/public/AUTH.mdAgent-facing WorkOS auth reference, synced from nyuchi/api-gateway.
/.well-known/mcp/server-card.jsonsite/public/.well-known/mcp/MCP server card for the nyuchi-docs-mcp worker at /mcp.
/.well-known/security.txtsite/src/worker/security-txt.tsRFC 9116 disclosure contact — generated per request, not a static file.

security.txt is dynamic because RFC 9116 makes Expires mandatory and caps it under one year, so a checked-in file silently becomes non-compliant as it ages. site/wrangler.toml sets run_worker_first = true, so every request already passes through site/src/worker/gate.ts; it answers this path before the gate check and before the asset router, deriving Expires from the request time (180 days out). Same approach as nyuchi/nhimbe and nyuchi/kweli.

Why pnpm workspace

The search package (nyuchi-docs-search) is consumed by both nyuchi-docs (this repo, via workspace:*) and bundu-docs (separate repo, via the npm registry). Keeping it in the same workspace as the docs site means local changes to the search UI are picked up instantly during pnpm dev, while the published package is a single pnpm publish away.