Loki Mode Integration

Loki Mode is a multi-agent autonomous coding system that deploys up to 41 specialized agents — frontend, backend, testing, documentation, DevOps, and more — working in parallel on a single project. Without Dashcam, observing this swarm of agents means watching raw logs scroll past or checking state files manually. With Dashcam, every agent gets its own swim lane on the timeline, RARV cycle boundaries are marked as decision events, and quality gate results land as colored markers you can click to inspect.

Bridge Architecture

The Loki Mode integration uses a bridge — a daemon module that watches Loki Mode's state files and connects to its WebSocket server. The bridge is started when you launch Dashcam with the --loki flag:

dashcam start --loki

The bridge operates on two channels:

File watcher — Monitors the .loki/ directory in your project root for state changes:

• .loki/state/orchestrator.json — Phase transitions (planning, building, testing, deploying)
• .loki/queue/pending.json — Tasks waiting for agent assignment
• .loki/queue/completed.json — Finished tasks with results
• .loki/memory/episodic/ — Agent memory and context data

WebSocket client — Connects to Loki Mode's dashboard WebSocket on port 57374 for real-time event streaming. This gives Dashcam the same live feed that the Loki Mode dashboard uses, without duplicating any infrastructure.

.loki/state/ ──file watch──▶ Loki Bridge ──normalize──▶ SQLite + WebSocket
                                  ▲
ws://localhost:57374 ─────────────┘

41 Agent Swim Lanes

Loki Mode defines 41 agent types, each with a specific role. Dashcam maps every agent type to its own swim lane in the timeline. When you open a Loki Mode session, the timeline shows a dense, multi-lane visualization of parallel work:

┌─────────────────────────────────────────────────────────────┐
│ 09:00       09:15       09:30       09:45       10:00       │
├─────────────────────────────────────────────────────────────┤
│ orchestrator  ──◆──────────◆──────────◆──────────◆──       │
│ frontend      ────●──●──●──●──●──●──●──────────────        │
│ backend       ──────●──●──●──●──●──●──●──●──               │
│ api-design    ──●──●──●──────────────────────────────       │
│ test-writer   ────────────────●──●──●──●──●──●──●──        │
│ reviewer      ──────────────────────◆──◆──◆──◆──           │
│ ...           (35 more lanes as needed)                      │
└─────────────────────────────────────────────────────────────┘

Lanes only appear when an agent has events — if only 8 of the 41 agent types were active in a session, the timeline shows 8 lanes. Inactive agent types are hidden.

RARV Cycle Markers

Loki Mode agents operate in RARV cycles: Reason, Act, Reflect, Verify. The bridge detects these cycle boundaries and inserts them into the timeline as distinct event types:

When you scrub the playhead to a RARV marker, the Chat panel shows the agent's reasoning or reflection at that moment. The Code panel shows what changed during the Act phase. The Result panel shows what the application looked like after Verify.

This turns the opaque "autonomous agent loop" into a legible sequence of deliberate decisions.

Quality Gate Results

Loki Mode runs quality gates at defined checkpoints — code review, test coverage, performance benchmarks, accessibility audits. The bridge captures these results as decision events on the timeline:

• Passed gates appear as green check markers
• Failed gates appear as red X markers
• Warning gates appear as amber diamond markers

Click any quality gate marker to see the full gate report in the Decision panel: what was checked, what passed, what failed, and what the agent decided to do about it.

Phase Change Tracking

The bridge monitors .loki/state/orchestrator.json for phase transitions. When the orchestrator moves from planning to building, or from building to testing, a milestone marker appears on the orchestrator's swim lane.

Phase changes are valuable landmarks in a long session. They tell you where the major transitions happened without requiring you to read through hundreds of individual events.