Playground

The hosted kitchen-sink playground is the separate antiartificial/context-kitchen-sink app. It is the interactive demo surface for scenarios, source credibility, contradiction workflows, agent memory, auditor epistemics, DSL exploration, and current system checks.

This page is the companion CLI checklist for trying the same major contextdb surfaces from a local server.

Start a local server

go run ./cmd/contextdb \
  --rest-addr :7701 \
  --observe-addr :7702

Open the admin dashboard:

open http://localhost:7702/admin/

Health and feature surface

curl http://localhost:7701/v1/ping
curl http://localhost:7701/v1/version
curl http://localhost:7701/v1/features
curl http://localhost:7702/admin/api/metrics

Write and retrieve

curl -X POST http://localhost:7701/v1/namespaces/playground/write \
  -H "Content-Type: application/json" \
  -d '{
    "content": "contextdb combines vector search, graph evidence, source trust, recency, and utility.",
    "source_id": "playground/docs",
    "labels": ["Claim", "Playground"],
    "confidence": 0.92
  }'

curl -X POST http://localhost:7701/v1/namespaces/playground/retrieve \
  -H "Content-Type: application/json" \
  -d '{
    "text": "How does contextdb rank knowledge?",
    "top_k": 5,
    "score_params": {
      "similarity_weight": 0.4,
      "confidence_weight": 0.35,
      "recency_weight": 0.15,
      "utility_weight": 0.1
    }
  }'

Explain ranking

After writing at least two related nodes, compare their ranking factors:

curl -X POST http://localhost:7701/v1/namespaces/playground/rank/explain \
  -H "Content-Type: application/json" \
  -d '{
    "node_id": "NODE_UUID",
    "other_node_id": "OTHER_NODE_UUID",
    "text": "How does contextdb rank knowledge?",
    "max_depth": 2
  }'

The response includes a summary, margin, score-factor deltas, and graph evidence when available.

Ranking evaluation

contextdb eval ranking --out ranking-eval.json --report
contextdb eval ranking --markdown-out ranking-eval.md
contextdb eval ranking --compare ranking-eval.json --diff-markdown

In the admin UI, open Ranking Evaluation, choose a Top K, and optionally load a previous ranking-eval-v*.json baseline for an in-browser delta.

Knowledge acquisition dry-run

curl -X POST http://localhost:7701/v1/namespaces/playground/acquisition/plan \
  -H "Content-Type: application/json" \
  -d '{"budget": 5, "max_gaps": 3}'

curl -X POST http://localhost:7701/v1/namespaces/playground/acquisition/execute \
  -H "Content-Type: application/json" \
  -d '{
    "budget": 2,
    "max_results": 3,
    "allowed_source_ids": ["playground/docs"],
    "connectors": [
      {
        "id": "local-search-preview",
        "type": "search",
        "endpoint": "https://search.example.internal/contextdb",
        "allowed_source_ids": ["playground/docs"],
        "default_labels": ["acquired", "playground"]
      }
    ]
  }'

Add "execute": true only after the dry-run payload is reviewed and the connector endpoint is real.

Provider-backed connector server

OPENAI_API_KEY=... \
XAI_API_KEY=... \
ANTHROPIC_API_KEY=... \
contextdb connectors serve \
  --addr :7780 \
  --providers openai,xai,anthropic \
  --allowed-domains docs.example.com,github.com

Then use an endpoint such as http://localhost:7780/openai/search in the acquisition execution connector list.

Receipts and retry guidance

curl "http://localhost:7701/v1/namespaces/playground/acquisition/receipts"
curl "http://localhost:7701/v1/namespaces/playground/acquisition/retry-candidates"
curl "http://localhost:7701/v1/namespaces/playground/acquisition/retry-recommendations"

Debugger workflow

Use the admin UI for the most complete debugger experience:

1. Open http://localhost:7702/admin/.
2. Search the playground namespace.
3. Inspect a node to view source context, confidence history, contradiction paths, and graph neighbors.
4. Add two search results to Explain Rank to compare score factors.

Release and docs checks

npm run docs:build
contextdb docs schema-catalog verify --report --annotations
contextdb eval ranking baseline manifest verify --manifest ranking-baseline-manifest.json --report
contextdb closure-bundle verify --manifest closure-bundle-manifest.json --report

These commands match the current reliability direction: local docs build, published schema catalog verification, ranking baseline artifact verification, and closure bundle verification.

Hosted kitchen sink status

The current playground repo now includes a System tab for:

• Current contextdb version and feature-surface status.
• Acquisition connector dry-run preview.
• Acquisition receipt and retry recommendation inspection.
• Explain-rank comparison over seeded auditor claims.

The full Svelte admin dashboard still lives in the main contextdb server on observe port 7702. That dashboard is the first-class place for ranking evaluation, metrics, source-trust timelines, contradiction-path summaries, graph/source context, and debugger compare workflows.

The older test_contextdb.py script in this repository is only a lightweight illustrative smoke script and does not cover current features such as acquisition connectors, receipt retries, ranking baselines, schema catalog verification, or the Svelte debugger.