Hyperfold AI

Agent Logs

Stream and analyze agent activity, decisions, and reasoning in real-time.

Overview

The hyperfold agent logs command provides real-time visibility into agent behavior. Watch negotiations unfold, understand pricing decisions, and debug issues as they happen.

Stream Logs

Watch agent activity in real-time:

# Stream live logs from an agent
$ hyperfold agent logs sales-bot-01

[2025-12-19 14:30:15] [INFO] Session started: sess_abc123
[2025-12-19 14:30:15] [INFO] Buyer: openai_gpt4_buyer | User: user_xyz
[2025-12-19 14:30:16] [ACP] Search request: "waterproof running shoes"
[2025-12-19 14:30:17] [DATA] Found 8 products (confidence: 0.91)
[2025-12-19 14:30:18] [ACP] Quote request: prod_aero_x2, offer=$120
[2025-12-19 14:30:19] [CALC] Floor: $92, List: $180, Competitor: $165
[2025-12-19 14:30:19] [DECISION] Counter-offer: $162 (10% gold discount)
[2025-12-19 14:30:20] [ACP] Response: counter_offer, price=162.00
[2025-12-19 14:30:45] [ACP] Accept received: price=162.00
[2025-12-19 14:30:46] [PAYMENT] Processing SPT: spt_live_xxx
[2025-12-19 14:30:47] [SUCCESS] Order created: order_789
[2025-12-19 14:30:47] [INFO] Session complete: revenue=$162.00

# Follow logs (like tail -f)
$ hyperfold agent logs sales-bot-01 --follow

# Show logs from all agents
$ hyperfold agent logs --all --follow

Log Levels

LevelDescription
DEBUGDetailed internal state (development only)
INFONormal operations, session events
WARNUnusual conditions, near-limit situations
ERRORFailures requiring attention

Session Logs

View complete logs for a specific negotiation session:

$ hyperfold agent logs sales-bot-01 --session="sess_abc123"

SESSION: sess_abc123
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Started:    2025-12-19T14:30:15Z
Duration:   32 seconds
Status:     completed
Outcome:    conversion

BUYER INFO:
  Agent:        openai_gpt4_buyer
  User:         user_xyz (Gold tier)
  History:      12 previous orders, $2,450 total

CONVERSATION:
  Turn 1: Search "waterproof running shoes"
          → 8 products returned

  Turn 2: Quote prod_aero_x2 @ $120
          → Counter: $162 (10% discount applied)

  Turn 3: Accept $162
          → Order created

PRICING DECISIONS:
  List price:     $180.00
  Buyer offer:    $120.00
  Floor price:    $92.00
  Final price:    $162.00
  Discount:       10% (Gold tier)
  Margin:         45%

PAYMENT:
  Method:         Stripe SPT
  Amount:         $162.00
  Status:         succeeded
  Charge ID:      ch_xyz123

Reasoning Logs

See the agent's chain of thought for pricing decisions:

$ hyperfold agent logs sales-bot-01 --reasoning

[14:30:18] [REASONING] Processing quote request...

  INPUT CONTEXT:
    Product: AeroRun X2 ($180 list)
    Buyer offer: $120
    Customer: Gold tier, 12 orders, $2,450 lifetime
    Inventory: 847 units (overstock)
    Competitor: $165 (Amazon)

  THOUGHT PROCESS:
    1. Buyer offered $120, which is 33% below list price
    2. My floor price is $92 (cost $80 × 1.15 margin)
    3. The offer is above floor, so it's technically acceptable
    4. However, I can do better for my margin
    5. Customer is Gold tier → 10% discount applies
    6. Competitor is at $165, so I should stay below that
    7. Optimal price: $162 (10% off, beats competitor by $3)

  DECISION: counter_offer
  PRICE: $162.00
  REASONING TO BUYER: "I can do $162 for you as a Gold member -
                       that's 10% off and beats Amazon's price."

# Verbose reasoning with all context
$ hyperfold agent logs sales-bot-01 --reasoning --verbose

# Reasoning for specific decision types
$ hyperfold agent logs sales-bot-01 --reasoning --filter="decision:reject"

Filtering & Search

Find specific logs with powerful filtering:

# Filter by log level
$ hyperfold agent logs sales-bot-01 --level=error
$ hyperfold agent logs sales-bot-01 --level=warn

# Filter by time range
$ hyperfold agent logs sales-bot-01 --since="1h"
$ hyperfold agent logs sales-bot-01 --since="2025-12-19T10:00:00Z"
$ hyperfold agent logs sales-bot-01 --until="2025-12-19T12:00:00Z"

# Filter by event type
$ hyperfold agent logs sales-bot-01 --filter="type:payment"
$ hyperfold agent logs sales-bot-01 --filter="type:acp"
$ hyperfold agent logs sales-bot-01 --filter="type:error"

# Search log content
$ hyperfold agent logs sales-bot-01 --search="floor price"
$ hyperfold agent logs sales-bot-01 --search="prod_aero_x2"

# Filter by outcome
$ hyperfold agent logs sales-bot-01 --filter="outcome:conversion"
$ hyperfold agent logs sales-bot-01 --filter="outcome:rejected"
$ hyperfold agent logs sales-bot-01 --filter="outcome:abandoned"

# Combine filters
$ hyperfold agent logs sales-bot-01 \
  --since="24h" \
  --filter="outcome:rejected" \
  --level=info

Export & Analysis

Export logs for offline analysis or external tools:

# Export logs to file
$ hyperfold agent logs sales-bot-01 --since="24h" --output=logs.json

> [Export] Fetching logs from last 24 hours...
> [Write] Writing 1,247 log entries to logs.json
✓ Exported to logs.json

# Export as CSV for analysis
$ hyperfold agent logs sales-bot-01 --since="7d" --format=csv --output=logs.csv

# Export session summaries
$ hyperfold agent logs sales-bot-01 --sessions --since="24h" --output=sessions.json

{
  "sessions": [
    {
      "session_id": "sess_abc123",
      "started_at": "2025-12-19T14:30:15Z",
      "duration_seconds": 32,
      "outcome": "conversion",
      "revenue": 162.00,
      "discount_percent": 10,
      "turns": 3
    }
  ],
  "summary": {
    "total_sessions": 127,
    "conversions": 42,
    "conversion_rate": 0.33,
    "total_revenue": 8450.00,
    "avg_discount": 12.3
  }
}

# Stream to external logging
$ hyperfold agent logs sales-bot-01 --follow | tee -a agent.log

# Send to BigQuery for analysis
$ hyperfold agent logs export --destination=bigquery --dataset=agent_logs