Tool · Models & evaluation / Conversations & archives

Conversation Target Tagger

Label who or what a message is about while keeping the conversation context that explains it.

Repository: CinvanaAI/conversation-target-tagger

Explore the repository ↗Try or inspect it ↓

Recorded synthetic example. Source with a complete offline example.

See the idea in action.

Three chronologically indexed messages with project:GardenWatch labels from the declared demo callback.

Example input

The supplied synthetic GardenWatch conversations.

Captured result

[
  {
    "conversation_id": "conversation-mapping",
    "conversation_title": "Workshop notes",
    "conversation_create_time": 90,
    "conversation_create_time_iso": "1970-01-01T00:01:30Z",
    "node_id": "old-1",
    "parent_id": null,
    "children_ids": [],
    "role": "user",
    "sender": "User",
    "create_time": 101,
    "create_time_iso": "1970-01-01T00:01:41Z",
    "content_type": "text",
    "content_parts": [
      "Put the soldering station in the workshop."
    ],
    "content_text": "Put the soldering station in the workshop.",
    "hidden": false,
    "model_slug": null,
    "status": null,
    "end_turn": null,
    "sequence_in_conversation": 1,
    "conversation_index": 2,
    "sort_timestamp": 101.0,
    "gidx": 1,
    "targets": [
      "place:Workshop"
    ]
  },
  {
    "conversation_id": "conversation-new",
    "conversation_title": "Garden sensor",
    "conversation_create_time": 100,
    "conversation_create_time_iso": "1970-01-01T00:01:40Z",
    "node_id": "new-1",
    "parent_id": null,
    "children_ids": [],
    "role": "user",
    "sender": "User",
    "create_time": 102,
    "create_time_iso": "1970-01-01T00:01:42Z",
    "content_type": "text",
    "content_parts": [
      "Could the GardenWatch project record soil moisture?"
    ],
    "content_text": "Could the GardenWatch project record soil moisture?",
    "hidden": false,
    "model_slug": null,
    "status": null,
    "end_turn": null,
    "sequence_in_conversation": 1,
    "conversation_index": 1,
    "sort_timestamp": 102.0,
    "gidx": 2,
    "targets": [
      "project:GardenWatch"
    ]
  },
  {
    "conversation_id": "conversation-new",
    "conversation_title": "Garden sensor",
    "conversation_create_time": 100,
    "conversation_create_time_iso": "1970-01-01T00:01:40Z",
    "node_id": "new-2",
    "parent_id": null,
    "children_ids": [],
    "role": "assistant",
    "sender": "Assistant",
    "create_time": 103,
    "create_time_iso": "1970-01-01T00:01:43Z",
    "content_type": "text",
    "content_parts": [
      "Yes, the sensor can publish a bounded reading."
    ],
    "content_text": "Yes, the sensor can publish a bounded reading.",
    "hidden": false,
    "model_slug": null,
    "status": null,
    "end_turn": null,
    "sequence_in_conversation": 2,
    "conversation_index": 1,
    "sort_timestamp": 103.0,
    "gidx": 3,
    "targets": [
      "project:GardenWatch"
    ]
  }
]

Try the example.

From the repository root, follow the dependency requirements in the README. This example uses supplied synthetic material.

python -m pip install -e .
python -m examples.offline_demo

Three chronologically indexed messages with project:GardenWatch labels from the declared demo callback.

Complete setup and instructions ↗

The interesting part.

Strict one-result-per-message alignment while retaining non-mainline, hidden, and empty exported nodes.

Public continuation

Where it came from.

A September 2025 chronological-target tagging script beside a ChatGPT export experiment.

September 2025 — Historical chronological target script. Source ↗

Separates normalization, batching, validation, transport and CLI; adds bounded context, local defaults and safe output.

Shared subject matter with other conversation tools does not establish code ancestry or plug-in compatibility.

Source ↗