Review Python methods as code: what comes in, what it calls and changes, what it returns — with tests, callers, and models beside it.
PyPI: threadline-review · Get started · Documentation · GitHub · Report a bug
Threadline opens your Python repository in a local browser, or writes it to one HTML file. Choose an endpoint, command, or method to read its code under a one-line summary of its inputs, the project functions it calls, what it changes, and what it returns. Called functions, related tests, callers, and model declarations open beside the code; the call map shows everything it reaches.
It reads source without importing or running the project being reviewed. That project's dependencies do not need to be installed.
Requires Python 3.12 or later. In your Python environment:
The PyPI package is threadline-review; the installed command is threadline.
Use --pre for the current beta releases.
python -m pip install --pre threadline-review
threadline review /path/to/your/python-projectYour browser opens the review automatically. If it does not, open the local address printed in the terminal. For virtual-environment setup, see the installation guide.
threadline review /path/to/your/python-project --output review.htmlThis writes a single HTML file, opens it, and exits. It needs no server, internet
connection, or Python installation to view. Search, the call map, method code,
tests, callers, models, and Git comparisons work from the embedded snapshot. Add --base HEAD
to include changes or --no-open to generate the file without opening a browser.
Regenerate the file after editing source; omit --output for a local server with
Refresh source. The HTML contains the reviewed source, including a baseline
when requested. Large exports take longer to prepare and are capped at 256 MiB; use
--source-root or --exclude to narrow them, or use the local server for very
large codebases.
| Part | What it shows |
|---|---|
| Summary | In, Calls, Changes, and Returns. Select an item to highlight it in the code or open the call. |
| Code | Only the selected method. Select a name to highlight every use; underlined calls open beside the code. |
| Beside the code | The opened call, test, or caller with its calling line marked, then Tests, Callers, and Models as source. |
| Call map | Project functions the method reaches, each with the line that calls it and tags such as if, later, or probably. |
Open → moves into a method; Back returns with your highlight and side code intact.
| Page | How to use it |
|---|---|
| Endpoints | Pick an HTTP verb tab, such as GET or POST, then an endpoint. |
| Commands & tasks | Select a CLI command or background task. |
| Modules & methods | Pick a module, filter its methods, then select one. |
Search is available across all three pages. Each link points to source evidence. Unresolved sources and effects remain visible as gaps rather than invented values. The usage guide explains model references, navigation, and the limits of static data flow.
Compare a working tree with a Git baseline:
threadline review /path/to/your/python-project --base HEADOpen Changes to select a changed method or compare its before/after source. Changed files, rename-only records, and edits outside methods remain visible when no callable body changes. Current callers, historical callers, and possible impact are shown separately. These are direct source relationships, not a complete impact analysis.
For source roots, exclusions, snapshot-pinned JSON queries, and evidence retrieval, see the usage guide.
Threadline uses static analysis. A call either resolves to one project function, is
tagged probably (a likely target), is a library call, or is tagged can't tell
(decided at runtime) — supported, possible, external, and unknown in JSON
output. Name highlighting matches names within a method; it is not a runtime value trace. Common injected,
inherited, and constructed receivers can have possible source candidates. Dynamic
dispatch can remain unresolved; the view does not establish what happened at runtime.
The header shows analysis problems, and Coverage lists parse failures and unmodeled
syntax.
This is a beta. Linux and macOS are tested with Python 3.12–3.14; Windows is experimental. Human reviewer sessions remain pending before a stable release.
- Capabilities and limits
- Validation and platform coverage
- Keyboard access and accessibility
- Changelog
See the contributor guide for development and testing, and the publishing guide for releases.