Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Nested lingtai-portal-guide reference for topology tape frames, replay chunks, portal API endpoints, and the Network JSON response.
.claude/skills/lingtai-ai-portal-guide-topology-and-api/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-03 | ✗→✓ | ▲ Improved | -50% | 0% |
| case-01 | ✗→✓ | ▲ Improved | -7% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -25% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -44% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -32% | 0% |
This is a nested lingtai-portal-guide reference. It covers the on-disk topology tape, replay chunks, HTTP endpoints, and live network JSON shape.
topology.jsonlEach line is a TapeFrame:
json{ "t": 1744567890123, "net": { "nodes": [...], "avatar_edges": [...], "contact_edges": [...], "mail_edges": [...], "stats": {...} } }
t is milliseconds since epoch. net is the same Network object /api/network serves; Network response shape below is a representative example of its core fields.
Historical data is stored as delta-encoded, keyframed, gzipped JSON chunks bucketed by hour under .lingtai/.portal/replay/chunks/. The manifest.json in that chunks directory lists all available chunks with their time ranges and frame counts.
All endpoints are served at http://localhost:<port>.
| Endpoint | Method | Response | Description | |----------|--------|----------|-------------| | /api/network | GET | Network JSON | Live network state (nodes, edges, stats) | | /api/topology | GET | JSON array of TapeFrame | Full topology tape (can be large) | | /api/topology/manifest | GET | ReplayManifest JSON | Chunk metadata for replay | | /api/topology/chunk?start=<hourMs> | GET | Chunk JSON (optionally gzip-encoded when requested) | Frames for one hour-bucket start time | | /api/topology/progress | GET | JSON object such as { "current": N, "total": M } (or {}) | Reconstruction progress parsed from reconstruct.progress | | /api/topology/rebuild | POST | ReplayManifest JSON | Trigger tape reconstruction from source data and rewrite replay chunks |
json{ "nodes": [ { "address": "orchestrator", "agent_name": "orchestrator", "nickname": "小灵", "state": "ACTIVE", "alive": true, "is_human": false, "capabilities": ["avatar", "search"] } ], "avatar_edges": [ {"parent": "orchestrator", "child": "avatar-1", "child_name": "avatar-1"} ], "contact_edges": [ {"owner": "orchestrator", "target": "human", "name": "human"} ], "mail_edges": [ {"sender": "orchestrator", "recipient": "human", "count": 5} ], "stats": { "active": 2, "idle": 1, "stuck": 0, "asleep": 0, "suspended": 3, "total_mails": 42 } }
Other measured skills in the registry, with their headline benchmark lift.