8495e4d549
More comprehensive handling in count_tokens(force_accurate==True): In case model is neither OpenAI nor Anthropic Claude, simply use an elbow room factor in order to force a more conservative estimate.
2025-03-24 15:47:35 +02:00
dd80276f3f
Support cloning repo
...
Support forcing accurate token calculation (claude)
Help docs: Add desired branch in case of user supplied git repo, with default set to "main"
Better documentation for getting canonical url parts
2025-03-23 09:55:58 +02:00
81dea65856
Format files by pre-commit run -a
...
Signed-off-by: Yu Ishikawa <yu-iskw@users.noreply.github.com >
2024-10-30 10:00:36 +09:00
84b80f792d
protections
2024-08-09 21:44:00 +03:00
9c3673209d
TokenEncoder
2024-04-03 08:42:50 +03:00
901c1dc3f0
issue tool
2023-09-06 08:43:01 +03:00
e0f295659d
A less hacky way
2023-08-09 12:17:54 +03:00
ed8554699b
bug fixes and updates
2023-08-03 16:05:46 -07:00
d7b77764c3
Support context aware settings (for each incoming request), support override of settings, refactor CLI to use pr_agent.py
2023-08-01 14:43:26 +03:00
02a1d8dbfc
Add support for fallback models
2023-07-23 16:16:36 +03:00
e2a4cd6b03
docstring
2023-07-20 10:51:21 +03:00
259fa84eeb
disabling encoding error on special_tokens
2023-07-06 13:22:12 +03:00
4b4d91dfe9
Initial commit - PR-Agent OSS release
2023-07-06 00:21:08 +03:00
