Comparing data across tools

Goal: Compare the data available in the JSON/transcripts of each tool supported by ai-footprint.

Update: This document is updated every time a new tool is added.

Legend

Symbol	Meaning
✅	Available in this tool
❌	Not available
⚠️	Partially available (different structure)

Comparison table

Field	Claude Code (JSONL)	Opencode (JSON)	Pi (JSONL)	Usage in ai-footprint
Identifiers
`sessionId`	✅	✅ `session_id`	✅ header `type:"session"`, `id`	Join key with the `sessions` table
`uuid` / `id`	✅ `msg_id`	✅ `message.id`	✅ `message.id` (entry `type:"message"`)	Unique key per message
Model
`model` (string)	✅ `"claude-opus-4-8"`	⚠️ `{providerID, modelID}`	✅ `message.model`	Normalized to `provider` + `model` in `InferenceEvent`
`provider`	❌ (inferred from `model`)	✅ `model.providerID`	✅ `message.provider`	Allows the EcoLogits lookup
Tokens
`input_tokens`	✅	✅ `tokens.input`	✅ `message.usage.input`	Impact calculation
`output_tokens`	✅	✅ `tokens.output`	✅ `message.usage.output`	Impact calculation
`cache_creation_input_tokens`	✅	⚠️ `tokens.cache.write`	✅ `message.usage.cacheWrite`	Impact calculation (cache)
`cache_read_input_tokens`	✅	⚠️ `tokens.cache.read`	✅ `message.usage.cacheRead`	Impact calculation (cache)
`reasoning_tokens`	❌	✅ `tokens.reasoning`	❌	To be validated with EcoLogits
Time
`timestamp` (ISO 8601)	✅	⚠️ Unix timestamp ms	✅ native ISO 8601	Converted to ISO 8601 UTC
`active_seconds` (delta)	✅ Computed	✅ `time.completed - time.created`	✅ Computed (delta between consecutive timestamps)	Active time measurement
Cost
`cost` (USD)	❌ (not read by the collector)	✅ `cost` (DB + exports)	⚠️ `message.usage.cost.total` (present but not stored)	Not stored — no `cost` field in `InferenceEvent`
`latency` (ms)	❌ (not read by the collector)	❌	❌	Not currently used
Session metadata
`title`	❌	✅ `session.title`	❌	Not stored (note: to be reconsidered)
`directory`	✅ `cwd`	✅ `session.directory`	✅ `session.cwd`	Used for `project` (basename)
`slug`	❌	✅ `session.slug`	❌	Not stored (note: to be reconsidered)
`path`	❌	✅ `session.path`	❌	Not stored (note: to be reconsidered)
Status
`role` (user/assistant)	✅	✅	✅ `message.role`	Filtering (only assistant messages count)
`error`	❌	✅ `error.name`, `error.data`	❌	Not stored — no `error` field in `InferenceEvent`
`archived`	❌	✅ `time_archived`	❌	Not currently used
Collection
Storage format	JSONL (line by line)	SQLite (local DB)	JSONL (line by line)	Tool-specific collectors
Trigger	`hooks.Stop` (end of session)	`session.idle` (plugin)	`session_shutdown` (extension)	Automatic installation
Backfilling	Direct JSONL read	Direct SQLite read	Direct JSONL read	Tool-specific backfilling scripts

Details per tool

Claude Code

Available sources:

Session transcripts: ~/.claude/projects/**/*.jsonl — the only source actually read by ClaudeCodeCollector (glob + parsing of all **/*.jsonl files).

Structure of an event (transcript):

{
  "type": "assistant",
  "message": {
    "model": "claude-opus-4-8",
    "usage": {
      "input_tokens": 8427,
      "output_tokens": 287,
      "cache_creation_input_tokens": 7052,
      "cache_read_input_tokens": 8020
    }
  },
  "sessionId": "sess-A",
  "uuid": "u1",
  "cwd": "/Users/me/DEV/projA",
  "timestamp": "2026-06-27T10:08:45.619Z"
}

Collector: ClaudeCodeCollector (direct JSONL read)

Opencode

Available sources:

Local DB: ~/.local/share/opencode/opencode.db (SQLite)
Plugin exports: ~/.ai-footprint/crush-exports/<sessionId>.json
SSE SDK: event.subscribe() (real time)

Structure of a session (session table):

{
  "id": "ses_120fa4dc7ffeSg6YofUeAPBX1P",
  "title": "New session - 2026-06-19T08:35:53.016Z",
  "directory": "/Users/renaudheluin/.agents",
  "model": "{\"id\":\"Qwen3.6-35B-A3B-4bit\",\"providerID\":\"myprovider\"}",
  "tokens_input": 145920,
  "tokens_output": 11950,
  "tokens_reasoning": 0,
  "tokens_cache_read": 0,
  "tokens_cache_write": 0,
  "time_created": 1781858206763,
  "time_updated": 1781858206763
}

Structure of a message (message table):

{
  "id": "msg_edf05b24e001gD1yQMa8Rqwfpa",
  "session_id": "ses_120fa4dc7ffeSg6YofUeAPBX1P",
  "data": "{\"role\":\"assistant\",\"time\":{\"created\":1781858153054,\"completed\":1781858153428},\"model\":{\"providerID\":\"myprovider\",\"modelID\":\"Qwen3.6-35B-A3B-4bit\"},\"tokens\":{\"input\":0,\"output\":0,\"reasoning\":0,\"cache\":{\"read\":0,\"write\":0}},\"cost\":0}"
}

Structure of a plugin export:

{
  "info": {
    "id": "sess-abc123",
    "slug": "mon-projet",
    "directory": "/home/user/project",
    "model": { "id": "claude-sonnet-4-20250514", "providerID": "anthropic" },
    "tokens": {
      "input": 8427,
      "output": 287,
      "reasoning": 0,
      "cache": { "read": 8020, "write": 7052 }
    },
    "time": { "created": 1719741600000, "updated": 1719742500000 }
  },
  "messages": [
    {
      "info": {
        "id": "msg-1",
        "role": "assistant",
        "time": { "created": 1719741700000, "completed": 1719741800000 },
        "model": {
          "id": "claude-sonnet-4-20250514",
          "providerID": "anthropic"
        },
        "tokens": {
          "input": 8427,
          "output": 287,
          "reasoning": 0,
          "cache": { "read": 8020, "write": 7052 }
        },
        "cost": 0.00123
      }
    }
  ]
}

Collector: CrushCollector (DB backfilling + plugin export reading)

Pi

Available sources:

Session transcripts: ~/.pi/agent/sessions/--<cwd>--/<timestamp>_<uuid>.jsonl

Structure of a session file (one type:"session" header entry, then type:"message" entries):

{"type":"session","id":"sess-A","timestamp":"2026-06-27T10:00:00.000Z","cwd":"/Users/me/DEV/projA"}
{"type":"message","id":"u1","timestamp":"2026-06-27T10:08:45.619Z","message":{"role":"assistant","provider":"anthropic","model":"claude-opus-4-8","usage":{"input":8427,"output":287,"cacheRead":8020,"cacheWrite":7052,"totalTokens":15786,"cost":{"total":0}}}}

Collector: PiCollector (direct JSONL read, root = directory or single file)

Notes for future tools

When a new tool is added, update:

This document (comparison table)
The corresponding collector in ai_footprint/collectors/
The tests in tests/test_<tool>_collector.py

Required fields to map into InferenceEvent:

provider (string)
model (string)
input_tokens (int)
output_tokens (int)
cache_creation_tokens (int)
cache_read_tokens (int)
timestamp (ISO 8601 UTC)
project (string, basename of the directory)
session_id (string)
msg_id (string)

Optional fields already mapped in InferenceEvent:

active_seconds (float)
client (string, identifies the tool)

Fields not yet mapped (present in at least one tool, absent from InferenceEvent — to be added if a product need arises):

reasoning_tokens (int)
cost (float, USD)
error (string)

Document created on 2026-06-30, updated on 2026-07-13. To be updated with each new tool added.