Smartgift-AI/AgentHub
14
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 103 Wiki Activity

fix(chat): count a subagent's edits in the turn's Edited N files card (#86)

Browse Source 
A skill that hands its work to subagents (vp-auto) showed "Edited 1 file"
for a turn that changed 4: a subagent's Write/Edit calls are in its own
agent-<id>.jsonl, so the card only saw the parent's.

- server/lib/agent-edits.js reads a subagent's transcript (readLines, a
  member's only as theirs) for the files it wrote or edited with their +/-
  lines, nested subagents folded in (3 levels), cached by size + mtime.
- The transcript route puts them on each Claude Agent/Task result as
  `edits`; the card adds them to the turn's own edits. A file only a
  subagent wrote is listed, never an artifact card (its versions are
  rebuilt from this transcript's inputs, which a subagent's are not).
This commit is contained in:
nntrivi2001
2026-10-23 14:26:29 +07:00
parent c8bc4a95a9
commit 5ccbaf5c23
7 changed files with 330 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/API.md
+1 -1
View File
@@ -2081,7 +2081,7 @@ GET /api/sessions/:id/transcript?agent_id=&run_id=&limit=&after=&before=&offset=
`transcripts` lists the session's conversation files — the main one first, then each subagent's and compaction's — as `{ transcripts: [{ id, name, type: "main"|"subagent"|"compaction", subagent_type?, db_agent_id, … }] }`; `db_agent_id` links an entry to its `agents` row (`null` when none matches). A Codex, Antigravity, Cursor, omp or OpenCode session lists its own tool's journal instead.
`transcript` reads one of them as structured messages: no `agent_id` = the main conversation, `agent_id` (+ `run_id` for a Workflow-tool agent) = a subagent's — each one plain name, never a path. `limit` defaults to 50 (max 200); `after=<line>` returns only messages past that JSONL line (the chat's tail), `before=<line>` the page before it (history), `offset` is the legacy pager. Answers `{ messages, total, has_more, last_line, first_line }` (empty when the file is gone). The live file under the Claude projects folder is preferred, else the snapshot taken at import. `404 NOT_FOUND` for an unknown session.
`transcript` reads one of them as structured messages: no `agent_id` = the main conversation, `agent_id` (+ `run_id` for a Workflow-tool agent) = a subagent's — each one plain name, never a path. `limit` defaults to 50 (max 200); `after=<line>` returns only messages past that JSONL line (the chat's tail), `before=<line>` the page before it (history), `offset` is the legacy pager. Answers `{ messages, total, has_more, last_line, first_line }` (empty when the file is gone). A Claude Agent/Task `tool_result` whose subagent (or one it started, up to 3 levels) wrote or edited files carries them as `edits: [{ path, added, removed, counted }]`, read from the subagent's own transcript (`server/lib/agent-edits.js`, cached by size + mtime) — the chat's "Edited N files" card counts them with the turn's own edits. The live file under the Claude projects folder is preferred, else the snapshot taken at import. `404 NOT_FOUND` for an unknown session.
---