[python/knowpro] Release to PyPI; support WebVTT; GMail extraction (#1603)

Main accomplishment is that we now publish to PyPI whenever a new tag is
pushed (to the branch).
Additionally, added support for ingesting .vtt files (for the MP demo).
Finally, add tool for extracting GMail messages (probably only usable by
me?).

Most of the code was written by Claude Sonnet 4 and 4.5, with occasional
help from GPT-5 (GMail extraction) and manual adjustments.

Author: gvanrossum-ms

.github/workflows/release-py.yml

@@ -0,0 +1,65 @@
+# Copyright (c) Microsoft Corporation.
+# Licensed under the MIT License.
+
+name: Release (PyPI via Trusted Publishing + uv)
+
+on:
+  push:
+    tags: [ "v*" ]      # tag to publish
+  workflow_dispatch:     # manual run
+
+permissions:
+  contents: read
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    defaults:
+      run:
+        shell: bash
+        working-directory: python/ta     # your project subdir
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Set up Python
+        uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+
+      - name: Install uv
+        run: |
+          curl -LsSf https://astral.sh/uv/install.sh | sh
+          echo "$HOME/.local/bin" >> "$GITHUB_PATH"
+
+      - name: Create .venv and install deps
+        run: uv sync
+
+      - name: Build sdist + wheel (via make)
+        run: make build                   # runs `uv build`, outputs to dist/
+
+      - name: Upload build artifacts
+        uses: actions/upload-artifact@v4
+        with:
+          name: dist
+          path: python/ta/dist/
+
+  publish:
+    needs: build
+    runs-on: ubuntu-latest
+    environment:
+      name: pypi
+      url: https://pypi.org/p/typeagent-py
+    permissions:
+      id-token: write     # REQUIRED for Trusted Publishing (no tokens!)
+      contents: read
+    steps:
+      - uses: actions/download-artifact@v4
+        with:
+          name: dist
+          path: dist
+
+      - name: Publish to PyPI
+        uses: pypa/gh-action-pypi-publish@release/v1
+        # For TestPyPI first, add:
+        # with:
+        #   repository-url: https://test.pypi.org/legacy/

python/ta/.github/copilot-instructions.md

@@ -8,3 +8,5 @@ venv
 /evals
 /junk
 __pycache__
+testdata/MP/
+db

python/ta/.gitignore

@@ -45,7 +45,7 @@ scaling: venv
 
 .PHONY: build
 build: venv
-	.venv/bin/python -m build --wheel
+	uv build
 
 .PHONY: venv
 venv: .venv

.copilot-instructions.md renamed to python/ta/AGENTS.md

@@ -10,45 +10,9 @@ This is an in-progress project aiming at a Pythonic translation of
 `TypeAgent/ts/packages/knowPro` and a few related packages to Python.
 (Pythonic because it uses Python conventions and types as appropriate.)
 
-### Original description
-
-(Not all of this matches the current state of the code.)
-
-- Python class names correspond 1:1 to TS interface or type names.
-  (Much has changed here.)
-- Field and method names are converted from camelCase to python_case.
-- I-named interfaces become `Protocol` classes.
-  (Or in some cases `ABC`s.)
-- Other interfaces and structured types become dataclasses.
-- Union types remain union types.
-- Except unions of string literals become `Literal` types.
-- Not all of the TS class hierarchy is translated literally.
-
-### How to build
-
-Tested on Mac and on Ubuntu 22 under WSL.
-Should work on most UNIXoids.
-(For Windows, look at the `make.bat` script.)
-
-- Install Python 3.12 or higher (get it from
-  [python.org](https://www.python.org/downloads/) for Mac,
-  or run `sudo apt install python3.12` for Linux)
-- Run `make all`
-- You should now have a wheel file under `dist/`
-- To install: `python3.12 -m pip install dist/typeagent-0.1.0-py3-none-any.whl`
-- TODO: Upload that wheel to PyPI
-- To clean up, run `make clean`
-
-### How to test
-
-- Set your environment to contain the necessary OpenAI or Azure API keys
-  (OPENAI_API_KEY or AZURE_OPENAI_API_KEY)
-- Run unit tests: `make test`
-- Review coverage with `make coverage`
-- Interactively testing queries: `make demo`
-- Comparing to a validated set of questions and expected answers:
-  - Obtain or construct a JSON file of q/a pairs and install in testdata
-  - Run `make compare` (takes about 5-10 seconds per q/a pair)
+### Documentation
+
+(To be written. Sorry.)
 
 ## Trademarks