Agus Zubiaga
8b5835de17
agent: Improve initial file search quality ( #29317 )
...
This PR significantly improves the quality of the initial file search
that occurs when the model doesn't yet know the full path to a file it
needs to read/edit.
Previously, the assertions in file_search often failed on main as the
model attempted to guess full file paths. On this branch, it reliably
calls `find_path` (previously `path_search`) before reading files.
After getting the model to find paths first, I noticed it would try
using `grep` instead of `path_search`. This motivated renaming
`path_search` to `find_path` (continuing the analogy to unix commands)
and adding system prompt instructions about proper tool selection.
Note: I know the command is just called `find`, but that seemed too
general.
In my eval runs, the `file_search` example improved from 40% ± 10% to
98% ± 2%. The only assertion I'm seeing occasionally fail is "glob
starts with `**` or project". We can probably add some instructions in
that regard.
Release Notes:
- N/A
2025-04-23 21:24:41 -03:00
..
2025-04-07 18:10:01 +00:00
2025-04-23 19:27:56 -03:00
2025-04-19 23:26:29 +00:00
2025-04-16 19:46:36 -04:00
2025-04-02 13:26:34 -07:00
2025-04-23 20:54:09 +00:00
2025-04-22 09:01:01 +00:00
2025-04-23 20:54:09 +00:00
2025-04-02 13:26:34 -07:00
2025-04-21 20:16:51 +00:00
2025-04-23 15:43:33 -03:00
2025-04-23 21:24:41 -03:00
2025-04-02 13:26:34 -07:00
2025-04-23 20:54:09 +00:00
2025-04-15 09:35:13 -06:00
2025-04-02 13:26:34 -07:00
2025-04-02 13:26:34 -07:00
2025-04-19 23:26:29 +00:00
2025-04-07 20:26:55 +00:00
2025-04-10 22:58:41 +00:00
2025-04-23 20:54:09 +00:00
2025-04-02 13:26:34 -07:00
2025-04-11 17:38:14 +00:00
2025-04-23 20:54:09 +00:00
2025-04-02 13:26:34 -07:00
2025-04-23 23:10:47 +00:00
2025-04-23 20:54:09 +00:00
2025-04-02 13:26:34 -07:00
2025-04-02 13:26:34 -07:00
2025-04-02 13:26:34 -07:00
2025-04-17 17:29:19 +00:00
2025-04-21 11:28:49 -03:00
2025-04-23 15:43:33 -03:00
2025-04-23 20:54:09 +00:00
2025-04-19 23:26:29 +00:00
2025-04-02 13:26:34 -07:00
2025-04-23 22:27:27 +00:00
2025-04-23 22:27:27 +00:00
2025-04-15 14:59:42 +00:00
2025-04-22 19:35:47 -04:00
2025-04-23 22:27:27 +00:00
2025-04-19 23:26:29 +00:00
2025-04-22 05:25:09 +00:00
2025-04-02 13:26:34 -07:00
2025-04-23 22:27:27 +00:00
2025-04-23 21:24:41 -03:00
2025-04-11 22:01:25 +00:00
2025-04-22 21:07:52 +00:00
2025-04-02 13:26:34 -07:00
2025-04-23 20:54:09 +00:00
2025-04-02 13:26:34 -07:00
2025-04-23 15:30:20 +00:00
2025-04-15 21:36:30 -04:00
2025-04-23 20:54:09 +00:00
2025-04-02 13:26:34 -07:00
2025-04-23 12:38:46 -04:00
2025-04-02 13:26:34 -07:00
2025-04-02 13:26:34 -07:00
2025-04-23 12:38:46 -04:00
2025-04-23 20:54:09 +00:00
2025-04-23 20:54:09 +00:00
2025-04-23 20:54:09 +00:00
2025-04-22 09:37:12 +00:00
2025-04-23 15:43:33 -03:00
2025-04-20 00:53:30 +00:00
2025-04-02 13:26:34 -07:00
2025-04-02 13:26:34 -07:00
2025-04-02 13:26:34 -07:00
2025-04-02 13:26:34 -07:00
2025-04-22 09:01:01 +00:00
2025-04-23 20:54:09 +00:00
2025-04-10 21:16:43 +00:00
2025-04-18 14:15:19 +00:00
2025-04-22 21:16:54 +00:00
2025-04-11 01:47:40 +05:30
2025-04-23 20:54:09 +00:00
2025-04-23 22:27:27 +00:00
2025-04-02 13:26:34 -07:00
2025-04-23 15:30:20 +00:00
2025-04-21 01:12:57 +00:00
2025-04-23 23:10:47 +00:00
2025-04-02 13:26:34 -07:00
2025-04-19 22:23:03 +08:00
2025-04-23 22:27:27 +00:00
2025-04-02 13:26:34 -07:00
2025-04-08 16:01:27 +00:00
2025-04-02 13:26:34 -07:00
2025-04-23 22:27:27 +00:00
2025-04-23 15:17:27 +00:00
2025-04-21 19:29:21 -04:00
2025-04-07 18:46:09 +00:00
2025-04-02 13:26:34 -07:00
2025-04-23 21:24:41 -03:00
2025-04-19 23:29:36 +00:00
2025-04-23 12:38:46 -04:00
2025-04-10 21:16:43 +00:00
2025-04-08 16:09:06 -06:00
2025-04-02 13:26:34 -07:00
2025-04-19 23:26:29 +00:00
2025-04-23 12:38:46 -04:00
2025-04-23 20:54:09 +00:00
2025-04-02 13:26:34 -07:00
2025-04-23 20:54:09 +00:00
2025-04-11 23:02:50 +00:00
2025-04-02 13:26:34 -07:00
2025-04-23 22:27:27 +00:00
2025-04-23 20:54:09 +00:00
2025-04-02 13:26:34 -07:00
2025-04-21 20:16:51 +00:00
2025-04-21 20:16:51 +00:00
2025-04-23 23:10:47 +00:00
2025-04-23 20:54:09 +00:00
2025-04-02 13:26:34 -07:00
2025-04-22 21:32:32 +00:00
2025-04-02 13:26:34 -07:00
2025-04-22 21:32:32 +00:00
2025-04-23 20:54:09 +00:00
2025-04-02 13:26:34 -07:00
2025-04-02 13:26:34 -07:00
2025-04-03 23:32:50 +02:00
2025-04-16 20:22:44 +00:00
2025-04-02 13:26:34 -07:00
2025-04-20 00:53:30 +00:00
2025-04-23 18:27:46 -04:00
2025-04-02 13:26:34 -07:00
2025-04-02 13:26:34 -07:00
2025-04-23 20:54:09 +00:00
2025-04-23 20:54:09 +00:00
2025-04-02 13:26:34 -07:00
2025-04-18 12:11:54 +00:00
2025-04-02 13:26:34 -07:00
2025-04-02 13:26:34 -07:00
2025-04-02 13:26:34 -07:00
2025-04-02 13:26:34 -07:00
2025-04-08 16:09:06 -06:00
2025-04-02 13:26:34 -07:00
2025-04-02 13:26:34 -07:00
2025-04-02 13:26:34 -07:00
2025-04-02 13:26:34 -07:00
2025-04-08 15:57:36 -06:00
2025-04-23 15:43:33 -03:00
2025-04-21 16:00:03 +00:00
2025-04-02 13:26:34 -07:00
2025-04-21 21:30:21 -07:00
2025-04-23 20:54:09 +00:00
2025-04-23 19:36:58 +03:00
2025-04-09 01:53:54 +00:00
2025-04-23 20:54:09 +00:00
2025-04-02 13:26:34 -07:00
2025-04-23 14:06:58 +00:00
2025-04-02 13:26:34 -07:00
2025-04-02 13:26:34 -07:00
2025-04-18 04:23:56 +05:30
2025-04-16 19:05:57 +02:00
2025-04-23 19:27:56 -03:00
2025-04-08 16:09:06 -06:00
2025-04-08 16:09:06 -06:00
2025-04-07 19:03:24 +00:00
2025-04-18 15:04:26 -06:00
2025-04-02 13:26:34 -07:00
2025-04-23 20:54:09 +00:00
2025-04-23 20:54:09 +00:00
2025-04-23 15:30:20 +00:00
2025-04-23 15:30:20 +00:00
2025-04-23 20:54:09 +00:00
2025-04-23 20:54:09 +00:00
2025-04-23 20:54:09 +00:00
2025-04-23 22:27:27 +00:00
2025-04-21 20:16:51 +00:00
2025-04-21 21:30:21 -07:00
2025-04-22 18:54:56 -04:00
2025-04-23 20:54:09 +00:00