freshcontext-mcp 0.3.17 → 0.3.19

package/docs/API_DESIGN.md

@@ -0,0 +1,434 @@
+# FreshContext REST API Design
+
+Status: design only
+
+FreshContext REST is a future host around FreshContext Core. It should make the Core pipeline easy to use over HTTP without moving runtime, cache, storage, billing, dashboard, or adapter behavior into Core.
+
+## Purpose
+
+FreshContext turns raw retrieval results into freshness-ranked, provenance-aware context for agents.
+
+The REST host should expose the simplest useful path:
+
+1. User has raw retrieved signals.
+2. User sends those signals to FreshContext.
+3. Core evaluates freshness, confidence, ranking, explanation, optional envelope, and optional provenance.
+4. Host returns ranked context.
+5. Agent or app uses best context first.
+
+## Product Spine
+
+```text
+raw signals
+  -> FreshContext Core evaluation
+  -> freshness-ranked, explained, provenance-aware context
+  -> agent / app / workflow
+```
+
+REST should wrap Core. It should not become a crawler, dashboard, cache, Store, billing system, or Worker replacement.
+
+## Endpoint Table
+
+| Method | Path | Purpose | Core function |
+|---|---|---|---|
+| POST | `/v1/evaluate` | Evaluate one signal | `evaluateSignal` |
+| POST | `/v1/evaluate-batch` | Evaluate and rank multiple signals | `evaluateSignals` |
+| POST | `/v1/stamp` | Produce FreshContext envelope text and JSON | `evaluateSignal` with `includeEnvelope: true` |
+| GET | `/v1/health` | Return host health/version | host-only |
+| GET | `/v1/spec` | Return static spec metadata and docs links | host-only |
+
+## POST /v1/evaluate
+
+Evaluates one signal.
+
+### Request
+
+```json
+{
+  "signal": {
+    "id": "sig_001",
+    "source": "https://example.com/article",
+    "source_type": "blog",
+    "title": "Example retrieved result",
+    "content": "Raw retrieved content...",
+    "published_at": "2026-05-24T12:00:00.000Z",
+    "retrieved_at": "2026-05-24T13:00:00.000Z",
+    "semantic_score": 0.84,
+    "date_confidence": "high",
+    "status": "success",
+    "metadata": {
+      "query": "browser agents"
+    }
+  },
+  "options": {
+    "includeEnvelope": true,
+    "includeProvenance": false
+  }
+}
+```
+
+### Response
+
+Shape: `CoreSignalEvaluationResult`.
+
+```json
+{
+  "signal": {
+    "contract_version": "freshcontext.signal.v1",
+    "id": "sig_001",
+    "source": "https://example.com/article",
+    "source_type": "blog",
+    "title": "Example retrieved result",
+    "content": "Raw retrieved content...",
+    "published_at": "2026-05-24T12:00:00.000Z",
+    "retrieved_at": "2026-05-24T13:00:00.000Z",
+    "semantic_score": 0.84,
+    "date_confidence": "high",
+    "status": "success",
+    "metadata": {
+      "query": "browser agents"
+    },
+    "reasons": []
+  },
+  "freshness_score": 98,
+  "utility": {
+    "score": 83.6,
+    "contextualRelevance": 84,
+    "decayFactor": 0.995,
+    "dateConfidenceFactor": 1,
+    "statusFactor": 1,
+    "lambda": 0.001,
+    "ageHours": 1,
+    "status": "success",
+    "reasons": []
+  },
+  "ranked": {
+    "id": "sig_001",
+    "source": "https://example.com/article",
+    "source_type": "blog",
+    "title": "Example retrieved result",
+    "content": "Raw retrieved content...",
+    "published_at": "2026-05-24T12:00:00.000Z",
+    "retrieved_at": "2026-05-24T13:00:00.000Z",
+    "semantic_score": 0.84,
+    "date_confidence": "high",
+    "status": "success",
+    "metadata": {
+      "query": "browser agents"
+    },
+    "freshness_score": 98,
+    "final_score": 0.882,
+    "confidence": "high",
+    "reason": "Strong semantic match and current freshness for blog."
+  },
+  "explanation": "Strong semantic match and current freshness for blog.",
+  "envelope": {
+    "context": {},
+    "text": "[FRESHCONTEXT]...",
+    "structured": {}
+  },
+  "reasons": []
+}
+```
+
+The REST host must not fetch upstream data, cache results, write D1, enforce Ha-Pri, or alter ranking policy.
+
+## POST /v1/evaluate-batch
+
+Evaluates and ranks multiple signals.
+
+### Request
+
+```json
+{
+  "signals": [
+    {
+      "id": "sig_a",
+      "source": "https://example.com/fresh",
+      "source_type": "blog",
+      "content": "Fresh relevant content...",
+      "published_at": "2026-05-24T12:00:00.000Z",
+      "retrieved_at": "2026-05-24T13:00:00.000Z",
+      "semantic_score": 0.8
+    },
+    {
+      "id": "sig_b",
+      "source": "https://example.com/old",
+      "source_type": "blog",
+      "content": "Older relevant content...",
+      "published_at": "2025-01-01T00:00:00.000Z",
+      "retrieved_at": "2026-05-24T13:00:00.000Z",
+      "semantic_score": 0.9
+    }
+  ],
+  "options": {
+    "includeEnvelope": false,
+    "includeProvenance": false
+  }
+}
+```
+
+### Response
+
+```json
+{
+  "evaluations": [
+    {
+      "signal": {},
+      "freshness_score": 98,
+      "utility": {},
+      "ranked": {
+        "final_score": 0.85
+      },
+      "explanation": "Strong semantic match and current freshness for blog.",
+      "reasons": []
+    }
+  ]
+}
+```
+
+`evaluate-batch` must use `evaluateSignals`. Ordering follows `ranked.final_score`, with stable tie ordering. `utility.score` is sidecar output and must not be used as default ordering.
+
+The REST host must not fetch upstream data, add host-specific scoring, replace Core ranking, or silently enable utility-weighted ranking.
+
+## POST /v1/stamp
+
+Produces a FreshContext envelope and structured JSON for one result.
+
+### Request
+
+```json
+{
+  "signal": {
+    "source": "https://example.com/source",
+    "source_type": "blog",
+    "content": "Raw content to wrap...",
+    "published_at": "2026-05-24T12:00:00.000Z",
+    "retrieved_at": "2026-05-24T13:00:00.000Z",
+    "semantic_score": 0.75,
+    "date_confidence": "high"
+  },
+  "options": {
+    "envelopeMaxLength": 8000,
+    "envelopeFormat": {
+      "publishedLabel": "Published",
+      "unknownDateText": "Publish date: unknown"
+    }
+  }
+}
+```
+
+### Response
+
+```json
+{
+  "context": {
+    "content": "Raw content to wrap...",
+    "source_url": "https://example.com/source",
+    "content_date": "2026-05-24T12:00:00.000Z",
+    "retrieved_at": "2026-05-24T13:00:00.000Z",
+    "freshness_confidence": "high",
+    "freshness_score": 98,
+    "adapter": "blog"
+  },
+  "text": "[FRESHCONTEXT]...",
+  "structured": {
+    "freshcontext": {
+      "source_url": "https://example.com/source",
+      "content_date": "2026-05-24T12:00:00.000Z",
+      "retrieved_at": "2026-05-24T13:00:00.000Z",
+      "freshness_confidence": "high",
+      "freshness_score": 98,
+      "adapter": "blog"
+    },
+    "content": "Raw content to wrap..."
+  }
+}
+```
+
+Implementation should prefer `evaluateSignal(..., { includeEnvelope: true })`.
+
+This endpoint must not become a cache metadata endpoint. Cache status, cache age, TTL, key version, KV/D1 state, and Worker-specific cache metadata belong to hosts that own cache policy.
+
+## GET /v1/health
+
+Returns host health and version metadata.
+
+### Response
+
+```json
+{
+  "ok": true,
+  "service": "freshcontext-rest",
+  "version": "0.1.0",
+  "core_available": true
+}
+```
+
+The health endpoint must not expose secrets, tokens, environment values, private diagnostics, Cloudflare internals, or account metadata.
+
+## GET /v1/spec
+
+Returns static spec metadata and documentation links.
+
+### Response
+
+```json
+{
+  "service": "freshcontext-rest",
+  "spec_version": "1.2",
+  "signal_contract": "freshcontext.signal.v1",
+  "docs": {
+    "core_api": "/docs/CORE_API.md",
+    "signal_contract": "/docs/SIGNAL_CONTRACT.md",
+    "freshcontext_spec": "/FRESHCONTEXT_SPEC.md",
+    "methodology": "/METHODOLOGY.md"
+  }
+}
+```
+
+This endpoint must not become a dynamic registry, dashboard, account page, or source catalog.
+
+## Error Shape
+
+REST errors should use one stable JSON shape:
+
+```json
+{
+  "error": {
+    "code": "invalid_request",
+    "message": "Request body must include signal.",
+    "details": []
+  }
+}
+```
+
+Recommended initial codes:
+
+| HTTP | Code | Meaning |
+|---:|---|---|
+| 400 | `invalid_request` | Missing or malformed JSON/body fields |
+| 405 | `method_not_allowed` | Wrong method for endpoint |
+| 413 | `payload_too_large` | Request exceeds host limit |
+| 415 | `unsupported_media_type` | Non-JSON request body |
+| 500 | `internal_error` | Unexpected host error |
+
+Do not encode Core ranking/freshness uncertainty as HTTP errors. Missing timestamps, failed content, low confidence, and omitted provenance are valid evaluation outcomes and should appear in the Core result with reasons.
+
+## REST Non-Goals
+
+The first REST design does not include:
+
+- dashboard
+- auth
+- tenancy
+- billing
+- webhooks
+- D1 persistence
+- cache policy
+- feed replacement
+- production Ha-Pri enforcement
+- utility-weighted ranking
+- vector database
+- adapter fetching
+- crawler/scraper orchestration
+- Worker runtime migration
+
+## Host / Core Boundary
+
+Core owns:
+
+- signal normalization
+- timestamp/future-date/failure guards
+- freshness scoring
+- context utility sidecar
+- default rank/explain behavior
+- optional envelope generation
+- optional Ha-Pri v2 material preparation
+
+REST host owns:
+
+- HTTP routing
+- JSON parsing and response formatting
+- HTTP status codes
+- payload size limits
+- request IDs/logging policy
+- CORS policy if needed
+- documentation examples
+
+MCP host owns:
+
+- MCP tool schemas
+- reference adapter invocation
+- MCP response shape
+- client compatibility
+
+Worker runtime owns:
+
+- Cloudflare runtime behavior
+- MCP transport
+- KV cache policy
+- D1/feed/cron/rate limiting
+- cache metadata injection
+
+Adapters own:
+
+- fetching
+- source-specific parsing
+- raw source normalization before Core
+
+Ops Pulse owns:
+
+- runtime diagnostics
+- Cloudflare health checks
+- operational assays
+
+Trust Scanner owns:
+
+- repo/package/release integrity
+- public-claim checks
+- trust gate reporting
+
+## Security and Privacy Notes
+
+- Do not log full request bodies by default.
+- Do not return environment variables, tokens, API keys, Cloudflare metadata, or private diagnostics.
+- Treat submitted content as client data.
+- Keep provenance optional and explicit.
+- Do not claim Ha-Pri v2 production enforcement unless a future Store/read-time verification path is implemented.
+- Do not accept non-JSON request bodies for evaluation endpoints.
+- Apply conservative host-level request size limits before invoking Core.
+
+## Tests Needed Before Implementation
+
+Before adding REST route code, add tests for:
+
+- `POST /v1/evaluate` request/response fixtures
+- `POST /v1/evaluate-batch` ordering parity with `evaluateSignals`
+- invalid JSON and missing body errors
+- unsupported content type
+- payload-too-large handling
+- missing timestamp and future timestamp outcomes
+- failed content cannot rank as fresh/high confidence
+- envelope opt-in behavior
+- provenance opt-in behavior and missing-material reasons
+- no cache/D1/fetch side effects
+- stable error shape
+
+## Future Expansion
+
+Later phases may add:
+
+- auth
+- tenancy
+- billing
+- dashboard
+- webhooks
+- D1 persistence
+- cache policy
+- utility-weighted ranking as an explicit mode
+- production Ha-Pri enforcement
+- vector database integration
+- adapter fetching
+
+Each expansion should be opt-in and separately designed. The first REST host should remain a thin wrapper around Core evaluation.

package/docs/CODEX_MCP_USAGE.md

@@ -0,0 +1,116 @@
+# FreshContext MCP Usage with Codex
+
+This note documents the verified Codex-compatible MCP setup for the local FreshContext repository.
+
+## What Codex can use
+
+Codex can launch FreshContext as a local MCP server over stdio.
+
+The verified local server entrypoint is:
+
+```powershell
+& '<node-executable>' '<repo-root>\dist\server.js'
+```
+
+The MCP server exposes 22 tools: the front-door `evaluate_context` tool plus 21 read-only reference adapters. The local smoke test verifies the package version, server version, expected tool count, the generic context-evaluation path, and representative adapter calls.
+
+No credential is required for the local stdio smoke path.
+
+## Local stdio setup
+
+Prerequisites:
+
+- Node.js 20 or newer
+- Repository dependencies installed with `npm install`
+- Built server output at `dist/server.js`
+
+From the repository root:
+
+```powershell
+cd '<repo-root>'
+npm install
+npm run build
+npm run smoke:stdio
+```
+
+For a local Codex setup, use the same Node executable and built server path validated by the smoke test:
+
+```toml
+[mcp_servers.freshcontext]
+command = '<node-executable>'
+args = ['<repo-root>\dist\server.js']
+```
+
+A more portable variant is also valid when `node` is available on Codex's PATH:
+
+```toml
+[mcp_servers.freshcontext]
+command = "node"
+args = ['<repo-root>\dist\server.js']
+```
+
+Keep this configuration in the local Codex config file, not in the repository. Do not commit machine-local paths.
+
+## Remote /mcp setup
+
+The repository declares a remote Streamable HTTP MCP endpoint in `server.json` and the README:
+
+```text
+https://freshcontext-mcp.gimmanuel73.workers.dev/mcp
+```
+
+For clients that need a stdio bridge to a remote MCP endpoint, the README uses `mcp-remote`:
+
+```toml
+[mcp_servers.freshcontext_remote]
+command = "npx"
+args = ["-y", "mcp-remote", "https://freshcontext-mcp.gimmanuel73.workers.dev/mcp"]
+```
+
+This remote path was identified from repository metadata. The validation in this task verified local stdio only, not remote Codex compatibility, Worker availability, or Codex Cloud support.
+
+## Verification steps
+
+Run the local smoke test:
+
+```powershell
+cd '<repo-root>'
+npm run smoke:stdio
+```
+
+Expected result:
+
+```json
+{
+  "ok": true,
+  "package_version": "0.3.19",
+  "server_version": "0.3.19",
+  "tool_count": 22
+}
+```
+
+Run whitespace validation before committing docs:
+
+```powershell
+git diff --check
+```
+
+Expected result: no output and exit code 0.
+
+## Safety notes
+
+- Do not place secrets, credentials, registry tokens, npm tokens, GitHub tokens, or Cloudflare tokens in Codex MCP config.
+- Do not read, edit, print, or commit local token files, local environment files, registry credentials, Cloudflare local state, or Wrangler state.
+- Do not commit local Codex config or machine-specific paths.
+- Prefer the local stdio path for this compatibility check because it is verified by `npm run smoke:stdio`.
+- Do not claim Codex Cloud support unless it is separately tested and documented.
+
+## Troubleshooting
+
+If Codex cannot start the server:
+
+- Confirm `dist/server.js` exists. If not, run `npm run build`.
+- Confirm Node is installed with `node -v`. The package requires Node.js 20 or newer.
+- If `node` is not found by Codex, use the full executable path from `node -p "process.execPath"`.
+- Run `npm run smoke:stdio` from the repository root and confirm `tool_count` is 22.
+- If the remote setup fails, verify network access, `npx` availability, and the remote endpoint separately. Do not treat remote failure as evidence that local stdio is broken.