lurqrun 0.0.1 → 0.0.3

package/README.md

@@ -8,93 +8,33 @@ lurq is a **companion to your coding agent**: it recommends and explains package
 
 ---
 
-## quick start (dev)
+## quick start: connect your agent
 
-```bash
-npm install
-cp .env.example .env       
-docker compose up -d       
-npm run build
-npx lurqrun --help
-```
+> **not public yet.** lurq unlocks at launch. [join the waitlist](https://lurq.run); the steps below are how it'll work once it's live.
 
-## development
+lurq is a **hosted service**: you don't run a database or a sync. get an API key,
+then run the guided installer:
 
 ```bash
-npm run dev -- <command>   
-npm test                    
-npm run typecheck         
-npm run lint               
+npx lurqrun install
 ```
 
-## configuration
-
-see [`.env.example`](./.env.example). only `DATABASE_URL` is required;
-gitHub/openAI keys allow full access to lurq commands but are not strictly necessary.
-
-| var | required | purpose |
-|---|---|---|
-| `DATABASE_URL` | yes | postgres connection (pgvector-enabled) |
-| `GITHUB_TOKEN` | recommended | github signals (stars, cadence, issues, archived) |
-| `EMBEDDING_PROVIDER` / `EMBEDDING_API_KEY` | no | openai embeddings; falls back to a local embedder |
-| `SUMMARY_PROVIDER` / `SUMMARY_API_KEY` | no | llm usage guides; falls back to the npm description |
-
-## setup the index
-
-```bash
-npx lurqrun db migrate     # create schema (pgvector) + load the curated seed list
-npx lurqrun sync           # scores from public APIs (~2 min)
-```
-
-`sync` is idempotent and tolerant of single-source outages. run it on a schedule
-(default frequency: daily) to keep the index fresh.
-
-## cli usage
-
-```bash
-npx lurqrun recommend "a form library for react"    
-npx lurqrun recommend "debounce a function"         
-npx lurqrun evaluate zod                            
-npx lurqrun compare drizzle-orm prisma typeorm      
-npx lurqrun verify lodahs                            
-```
+it prompts for your key, validates it, detects your installed assistants
+(Claude Code, Cursor, Windsurf, VS Code/Copilot, Codex), and writes a keyed
+remote MCP entry:
+`{ "type": "http", "url": "https://api.lurq.run/mcp", "headers": { "Authorization": "Bearer …" } }`.
+**no database credentials ever touch your machine.** restart your agent afterward.
 
-add `--json` to any command for machine output; `--category` and `--min-confidence`
-filter `recommend`.
+## what your agent gets (mcp tools)
 
-## mcp tools
-
-`npx lurqrun serve` starts the MCP server (stdio). it exposes:
-
-- **`recommend`** — best current packages for a described need (≤5, scored, with confidence)
-- **`evaluate`** — full evidence read for one package (scores, advisories, usage guide)
-- **`compare`** — 2–5 packages ranked by health
-- **`verify`** — is a package real, healthy, and not risky? (anti-hallucination guard)
-- **`diagram`** — a reference-architecture mermaid diagram for a stack (optional)
-
-every response is compact and carries a `dataAsOf` timestamp.
-
-## install into your agent
-
-```bash
-npx lurqrun install-skill --agent claude-code   # or cursor | windsurf | copilot | codex | all
-```
-
-merges a `lurq` MCP server entry into the agent's config (never overwriting other
-config) and drops a short instructions file telling the agent when to call lurq.
-restart the agent afterward.
-
-## how it works
-
-```
-public APIs (npm, github, deps.dev, bundlephobia)
-  → ingestion → scoring (+ confidence) → summaries/usage guides → embeddings
-  → postgres + pgvector
-  → { mcp server, cli }
-```
+once installed, the agent can call these tools over MCP. every response is
+compact and carries a `dataAsOf` timestamp.
 
-scores are computed from public signals — never hand-written. confidence
-labels (`proven` / `emerging` / `unproven`) reflect the strength of discovered evidence
+- **`recommend`**: best current packages for a described need (≤5, scored, with confidence)
+- **`evaluate`**: full evidence read for one package (scores, advisories, usage guide)
+- **`compare`**: 2 to 5 packages ranked by health
+- **`verify`**: is a package real, healthy, and not risky? (anti-hallucination guard)
+- **`diagram`**: a reference-architecture mermaid diagram for a stack (optional)
 
 ## outreach