Will you sell your GPT-4-32k API? If you would like to, please contact me.  #43

Activity

1. usemanusai commented on May 22, 2026

   

   usemanusai

   on May 22, 2026

   More actions

   EASY!

   Just do this:

   1. Go to:

   https://capy.ai?ref=876YSLPX

   2. Create an account

   3. Connect Capy to GitHub

   4. Select the correct GitHub (forked) repo

   5. Send a text prompt to the Capy AI chat something like:

   "From the github repo I need you to do a complete rewrite of gpt-repository-loader as a modular Python package, addressing all open issues and pull requests with a production-ready implementation."

   6. Wait about 10-20 minutes for Capy to complete

   7. Download new updated files from your forked repo

   8. Welcome to the world of AI (can be done for any issue / bug, just change the text prompt to what needs to be done)**

   This PR implements a complete rewrite of gpt-repository-loader as a modular Python package, addressing all open issues and pull requests with a production-ready implementation.

   Package & Tooling

   • Add pyproject.toml with [project] metadata, optional dependencies (git, tokens, clipboard, web), and entry points gpt-repository-loader / gpt-repo (Add pyproject.toml #63, 🚀 Add pipx installation support via pyproject.toml #64, Pipx installation #55).
   • Add .devcontainer/devcontainer.json for VS Code remote development (Create devcontainer.json #44).
   • Add GitHub Actions CI workflow testing Python 3.8–3.12 on Ubuntu, macOS, and Windows.
   • Remove legacy gpt_repository_loader.py script; refactor into gpt_repository_loader package with modules.

   Ignore Engine & Binary Detection

   • Implement full gitignore-style pattern matcher (ignore.py): negation patterns (!), leading slash anchoring, directory-only trailing slashes, ** path matching, character classes, comments, and escaping (Add negated-pattern support for .gptignore #46, ignore .gitignore along with .gptignore #24, ignore .tox #21, .gptignore should appropriately interpret leading slashes #40).
   • Merge .gitignore and .gptignore across the tree, with package-shipped fallback and CLI --exclude / --include overrides (ignore .gitignore along with .gptignore #24, adding a number of features to understand what is being ignored #52, Don't include non relevant files in the scan #56).
   • Detect binary files via extension allow/deny lists, magic-byte sniffing (PNG, JPEG, GIF, ELF, gzip, zip, etc.), and NUL byte heuristics (Ignore Binary Files #53, An option to compress the file contents #37, Ignore .DS_Store #42, added some essential terms to ignore #47).
   • Add --show-ignored and --json-report to list excluded files with matching patterns and reasons (adding a number of features to understand what is being ignored #52).

   Token Counting, Limits & Chunking

   • Integrate tiktoken for accurate token counting when installed; deterministic 4-char fallback otherwise (Add warning about token length being greater than GPT limits #26, Add a token limit flat to the command. #41).
   • Add --token-limit flag (per-model defaults via --model), warn when exceeded, and exit code 2 for CI (Add a token limit flat to the command. #41).
   • Implement --chunk to split oversized bundles into <output>.chunkNN.txt files (In many applications, output.text is too long to input #35).
   • Track per-file and total token counts, include stats in --preamble-stats and Web UI.

   Compression & Output Formats

   • Add per-file compression via --compress / --encoding-mode gzip+base64; reverse loader decodes transparently (An option to compress the file contents #37).
   • Support --format zip to produce a ZIP file containing both bundle.txt and raw file tree (gpt-4 doesn't seem to think this is better than a zip #48).
   • Support --format json to emit a JSON document with files, ignored list, warnings, and metadata.
   • Auto-upgrade to base64 encoding for files containing literal ---- or --END-- lines.

   Remote Repositories & Git Integration

   • Accept GitHub/GitLab/Bitbucket/Codeberg URLs, git@ SSH strings, and downloadable archives via url_loader.py; clone via GitPython or extract to temp dir (Support URLs #36, Use GitPython library #32, Add flag do the reverse process, write the output file(s) back to the repository #30).
   • Add --clone-depth and --branch for shallow clones.
   • Use GitPython for --tracked-only mode, repo introspection (HEAD SHA, branch, dirty flag), and git check-ignore integration (Use GitPython library #32, Don't include non relevant files in the scan #56).