[GH-ISSUE #104] Working with many and/or large files #64

New issue

Open

opened 2026-10-18 21:10:18 +00:00 by bsakaguchi · 2 comments

bsakaguchi commented

2026-10-18 21:10:18 +00:00

Owner

Originally created by @vaclcer on GitHub (Apr 3, 2026).
Original GitHub issue: https://github.com/open-webui/open-terminal/issues/104

I wanted to flag a couple of limitations I’ve run into when working with many or large files in Terminal. I understand these are edge cases and likely to be addressed over time, but sharing early feedback:

The LLM tends to call list_files frequently. Without pagination or safeguards (e.g., detecting large directories), this can easily overflow the model’s context window and fail silently.

Similarly, handling large files is problematic. Downloading a single large file doesn’t seem to stream; instead, it appears to load into memory first, which can cause the container to run out of memory (OOM) if RAM is insufficient.

I expect these scenarios will come up naturally as usage grows, but it might be worth explicitly considering cases with thousands of files and/or very large files (GB-scale).

Overall, great work on Open Terminal — really impressive so far. Thanks!

Originally created by @vaclcer on GitHub (Apr 3, 2026). Original GitHub issue: https://github.com/open-webui/open-terminal/issues/104 I wanted to flag a couple of limitations I’ve run into when working with many or large files in Terminal. I understand these are edge cases and likely to be addressed over time, but sharing early feedback: The LLM tends to call list_files frequently. Without pagination or safeguards (e.g., detecting large directories), this can easily overflow the model’s context window and fail silently. Similarly, handling large files is problematic. Downloading a single large file doesn’t seem to stream; instead, it appears to load into memory first, which can cause the container to run out of memory (OOM) if RAM is insufficient. I expect these scenarios will come up naturally as usage grows, but it might be worth explicitly considering cases with thousands of files and/or very large files (GB-scale). Overall, great work on Open Terminal — really impressive so far. Thanks!

bsakaguchi commented

2026-10-18 21:10:19 +00:00

Author

Owner

@nunodonato commented on GitHub (Apr 22, 2026):

Bumped into this as well. Was experimenting processing a 5Mb CSV file to generate some insights, and it just blew up the context window :(

@nunodonato commented on GitHub (Apr 22, 2026): Bumped into this as well. Was experimenting processing a 5Mb CSV file to generate some insights, and it just blew up the context window :(

bsakaguchi commented

2026-10-18 21:10:19 +00:00

Author

Owner

@dgarciaf18 commented on GitHub (Jul 29, 2026):

Suffering from this issue. Running read_file with large enough files will freeze the terminal and require an open-terminal server restart

@dgarciaf18 commented on GitHub (Jul 29, 2026): Suffering from this issue. Running read_file with large enough files will freeze the terminal and require an open-terminal server restart

bsakaguchi referenced this issue

2026-10-18 21:10:58 +00:00

[PR #64] [CLOSED] Rakgoale Ai Developer Services-image.yml #118