Skip to main content

Documentation Index

Fetch the complete documentation index at: https://condor.hummingbot.org/llms.txt

Use this file to discover all available pages before exploring further.

Each time you start a Trading Agent, it creates a new session that tracks all activity, decisions, and state changes.

​
Session Structure

Sessions are stored in the agent’s trading_sessions/ directory: 
trading_agents/my-strategy/
└── trading_sessions/
    ├── session_1/
    │   ├── journal.md
    │   └── snapshots/
    ├── session_2/
    │   ├── journal.md
    │   └── snapshots/
    └── session_3/
        ├── journal.md
        └── snapshots/

​
journal.md

The journal is the agent’s working memory for the session. It contains:

​
Summary Section

High-level overview updated each tick: 
## Summary

- **Session Start**: 2026-03-27 14:00 UTC
- **Current Tick**: 47
- **Active Executors**: 2
- **Session P&L**: +$127.50
- **Status**: Running

​
Decisions Section

Key decisions made during the session: 
## Decisions

### Tick 45 - 2026-03-27 15:32 UTC
**Decision**: Open long position on SOL-USDT
**Reasoning**: Price broke above 20-period high with increasing volume
**Action**: Created PositionExecutor (ID: exec_045)

### Tick 38 - 2026-03-27 15:10 UTC
**Decision**: Close grid executor
**Reasoning**: Volatility exceeded threshold, grid levels too tight
**Action**: Stopped GridExecutor (ID: exec_032)

​
Ticks Section

Record of each OODA loop iteration: 
## Ticks

### Tick 47 - 2026-03-27 15:35 UTC
- **Portfolio Value**: $10,127.50
- **Open Positions**: 1 (SOL-USDT long 10 @ 152.30)
- **Active Executors**: 2
- **Market Context**: SOL +2.3% 1h, funding 0.01%
- **Action Taken**: Monitoring, no changes

​
Executors Section

Active and completed executor states: 
## Executors

### Active
| ID | Type | Pair | Side | Entry | Current | P&L |
|----|------|------|------|-------|---------|-----|
| exec_045 | Position | SOL-USDT | BUY | 152.30 | 154.10 | +$18.00 |

### Completed This Session
| ID | Type | Close Type | P&L | Duration |
|----|------|------------|-----|----------|
| exec_032 | Grid | EARLY_STOP | +$45.20 | 2h 15m |

​
snapshots/

Full tick snapshots for debugging and replay. This is one of the most important parts of the system: observability. Each snapshot captures the complete state of a tick: 
{
  "tick": 47,
  "timestamp": "2026-03-27T15:35:00Z",
  "system_prompt": "...",
  "agent_response": {
    "text": "Trend flip bullish. All conditions met, opening long.",
    "decision": "create_executor",
    "reasoning": "Price within 2% of S1, EMAs aligned bullish"
  },
  "tool_calls": [
    {"tool": "run_routine", "args": {...}, "result": {...}},
    {"tool": "create_executor", "args": {...}, "result": {...}}
  ],
  "portfolio": {
    "total_value_quote": 10127.50,
    "positions": [...],
    "balances": {...}
  },
  "market_data": {
    "SOL-USDT": {
      "price": 154.10,
      "volume_24h": 1250000000,
      "funding_rate": 0.0001
    }
  },
  "executors": [...],
  "risk_state": {
    "daily_pnl": 127.50,
    "total_exposure": 1541.00,
    "is_blocked": false
  }
}
This lets you:
  • Debug exactly what the agent was thinking when it made a decision
  • Understand why a trade was placed or not placed
  • Replay the prompt that was passed to the agent
  • Audit all reasoning and tool calls
Retention: Maximum 100 snapshots per session. Older snapshots are pruned automatically.

​
Session Continuity

The ~/condor directory stores all agent state. Combined with ACP (Agent Client Protocol), this enables seamless continuity across interfaces:
InterfaceHow to Connect
Telegram/agent → Select agent → Resume session
Claude Codeclaude --agent ~/condor/trading_agents/my-strategy
Web DashboardSelect agent from dashboard (coming soon)
When you switch interfaces:
  • Same session continues
  • Full conversation history preserved
  • Agent state unchanged

​
Starting a New Session

To start a fresh session instead of resuming: Telegram: /agent → Select agent → New Session CLI: 
claude --agent ~/condor/trading_agents/my-strategy --new-session

​
Injecting Directives

You can send real-time instructions to a running agent using directives. The directive will be included in the agent’s next tick prompt under a USER DIRECTIVES section, then cleared. Telegram: /agent → Select running agent → Inject Directive Programmatic: 
engine.inject_directive("Reduce exposure, news event in 10 min.")
Common directive use cases:
  • Alert the agent to upcoming news events
  • Request exposure reduction before volatility
  • Override normal behavior temporarily
  • Provide market context the agent can’t observe
The agent sees the directive verbatim and decides how to act on it based on its strategy rules.

​
Session Lifecycle