Update PyPI releases to include sdist

.github/workflows/release.yml

@@ -93,21 +93,28 @@ jobs:
               raise SystemExit(f"unexpected import name: {src_py_lib.__name__}")
           PY
 
-      - name: Build wheel
+      - name: Build distributions
         id: build
         run: |
           dist_dir="build/release/dist"
           rm -rf build/release
           mkdir -p "${dist_dir}"
+          shopt -s nullglob
 
-          uv build --wheel --out-dir "${dist_dir}" --no-create-gitignore
+          uv build --wheel --sdist --out-dir "${dist_dir}" --no-create-gitignore
           project_wheels=("${dist_dir}"/*.whl)
           if [[ "${#project_wheels[@]}" -ne 1 ]]; then
             echo "::error title=Unexpected wheel count::Expected one project wheel, found ${#project_wheels[@]}."
             exit 1
           fi
+          source_distributions=("${dist_dir}"/*.tar.gz)
+          if [[ "${#source_distributions[@]}" -ne 1 ]]; then
+            echo "::error title=Unexpected source distribution count::Expected one source distribution, found ${#source_distributions[@]}."
+            exit 1
+          fi
           wheel_path="${project_wheels[0]}"
           wheel_name="$(basename "${wheel_path}")"
+          source_distribution_path="${source_distributions[0]}"
           checksum_path="${wheel_path}.sha256"
 
           (
@@ -117,6 +124,7 @@ jobs:
 
           echo "wheel_path=${wheel_path}" >> "${GITHUB_OUTPUT}"
           echo "wheel_name=${wheel_name}" >> "${GITHUB_OUTPUT}"
+          echo "source_distribution_path=${source_distribution_path}" >> "${GITHUB_OUTPUT}"
           echo "checksum_path=${checksum_path}" >> "${GITHUB_OUTPUT}"
 
       - name: Smoke test installed wheel
@@ -178,7 +186,9 @@ jobs:
         uses: actions/upload-artifact@v7
         with:
           name: pypi-distributions
-          path: ${{ steps.build.outputs.wheel_path }}
+          path: |
+            ${{ steps.build.outputs.wheel_path }}
+            ${{ steps.build.outputs.source_distribution_path }}
 
       - name: Publish GitHub release assets
         env:

README.md

@@ -1,67 +1,41 @@
 # src-py-lib
 
-Reusable libraries for Sourcegraph-adjacent Python projects
-
-This repo is the shared implementation layer for patterns which get
-rebuilt in separate scripts: API clients, HTTP retries/timeouts, structured logging,
-etc.
+Reusable libraries for Sourcegraph projects
 
 ## Experimental - This is not a supported Sourcegraph product
 
 This repo was created for Sourcegraph Implementation Engineering deployments,
 and is not intended, designed, built, or supported for use in any other scenario.
 Feel free to open issues or PRs, but responses are best effort.
 
-## Install from another project
+## Install
+
+From PyPI:
+
+```sh
+uv add src-py-lib
+```
+
+From this repository:
 
 ```sh
 uv add git+https://github.com/sourcegraph/src-py-lib.git
 ```
 
-## What is included
-
-- `src_py_lib.utils.logging` — centralized human stderr logs plus optional structured
-  JSONL events, run IDs, git commit metadata, context fields, event timing,
-  retention, startup metadata, and sanitized config snapshots.
-- `src_py_lib.utils.config` — Pydantic-backed `Config` models loaded from code
-  defaults, `python-dotenv` `.env` parsing, shell environment, and CLI
-  overrides, with typed values, required checks, safe snapshots, and `op://...`
-  reference resolution.
-- `src_py_lib.utils.http` — pooled `httpx` JSON HTTP client with a shared
-  30-second timeout, retry policy, `Retry-After` support, and contextual errors.
-- `src_py_lib.utils.tsv` — padded TSV writer for human-readable tabular exports,
-  with newline/tab cleanup, URL preservation, and Unicode-aware column widths.
-- `src_py_lib.clients.graphql` — shared GraphQL execution with automatic cursor
-  pagination, batched alias lookups, and schema introspection export.
-- `src_py_lib.clients.sourcegraph` — Sourcegraph GraphQL client with token
-  validation, endpoint normalization, connection streaming, and shared config
-  fields for `SRC_ENDPOINT` (default: `https://sourcegraph.com`) and
-  `SRC_ACCESS_TOKEN`.
-- `src_py_lib.clients.linear` — Linear GraphQL client with automatic cursor
-  handling, token validation, shared config fields, and injectable HTTP policy.
-- `src_py_lib.clients.slack` — Slack Web API client with token validation,
-  cursor pagination, and method pacing. Consider `slack_sdk` if usage grows
-  beyond simple GET, pagination, and rate-limit handling.
-- `src_py_lib.clients.github` — GitHub GraphQL client, PR URL parsing, and
-  batched PR lookups, with token validation. Defaults to `https://github.com`;
-  pass `github_url` for GitHub Enterprise Server. Keep lightweight for GraphQL;
-  GitHub SDKs help more for REST.
-- `src_py_lib.clients.one_password` — tiny 1Password CLI wrapper for signing in,
-  validating authenticated `op` access, and resolving `op://...` references after config loading.
-- `src_py_lib.clients.google_sheets` — Google Sheets API primitives with
-  spreadsheet access validation using gcloud Application Default Credentials or
-  a provided access token. Prefer Google's official libraries if Sheets usage
-  grows beyond small primitives, because auth, quota project, token refresh,
-  batching, and error shapes are subtle.
-
-Prefer this library for shared logging, HTTP policy, and thin API wrappers.
-Prefer vendor SDKs when they replace tricky auth, token refresh, retries,
-pagination, quota behavior, or complex request models.
-
-## Example
-
-Define one project-specific `Config` model, then load it once at CLI startup.
-For common CLI and client usage, import the curated root API:
+## Use it for
+
+- Typed config models loaded from defaults, `.env`, environment variables, and CLI flags
+- Root logger setup with terminal output, optional JSONL events, timing, and run metadata
+- A shared `httpx` JSON client with timeouts, retries, `Retry-After`, and contextual errors
+- Small helpers for TSV output, JSON caches, GraphQL pagination, and batched GraphQL queries
+- Thin clients for Sourcegraph, Linear, Slack, GitHub, Google Sheets, and 1Password
+
+Prefer vendor SDKs when they handle complex auth, token refresh, quota behavior,
+pagination, retries, or request models better than a thin wrapper
+
+## Quick example
+
+Define a project config, parse it once, and configure logging at startup:
 
 ```python
 from pathlib import Path
@@ -78,56 +52,22 @@ class LinearExportConfig(src.LinearClientConfig):
         help="Directory for generated files.",
     )
 
-config = src.parse_args(LinearExportConfig, description="Export Linear data.")
-client = src.linear_client_from_config(config)
-print(f"Writing files under {config.output_dir}")
-```
-
-Config precedence is: code defaults, `.env`, shell environment, then CLI
-overrides. API client modules can provide shared Config base classes such as
-`LinearClientConfig`, and `parse_args` resolves `op://...` references by
-default. `config_field(default=...)` supports aliases, store-true /
-store-false command flags, optional values, numeric bounds, and string patterns
-for simple CLIs. Pass a custom `argparse.ArgumentParser` to `parse_args` only when you
-need parsing beyond Config fields. Help text preserves description and
-argument-help newlines, and reserves enough option-column width for long config
-flags. Mark sensitive fields with `secret=True` so snapshots do not expose
-resolved values.
-
-## Logging example
-
-Configure logging once at process startup. Prefer configuring the root logger
-(`logger_name=""`, the default) so project modules and shared `src_py_lib` modules
-such as `src_py_lib.utils.http` are captured by the same terminal and JSONL handlers.
-Use `logging()` in CLIs to configure logging, add the command field to all
-structured events, and emit standard run/startup/run-end metadata.
-Use `debug()`, `info()`, `warning()`, `error()`, and `critical()` for one-off
-structured events. Use `event()` blocks around timed work; they emit `trace`,
-`span`, and nested `parent_span` fields. Use `start_level="debug"` to hide
-noisy start events while keeping end timing visible, and
-`omit_success_status=True` for very high-volume success events. Use `stage()`
-for workflow context such as `stage="apply"`.
-When the root logger is configured, noisy `httpx`/`httpcore` records are suppressed;
-`HTTPClient` emits structured `http_request` events instead.
-Run-end events include HTTP attempt/byte/status/retry counters. Set
-`LoggingSettings.resource_sample_interval_seconds` to emit DEBUG
-`resource_sample` events and include process resource totals on run end. Set
-`SRC_LOG_LEVEL=INFO` for a run to omit DEBUG events from the log file.
-`LoggingConfig` includes `--verbose/-v`, `--quiet/-q`, and `--silent/-s`
-shortcuts (also available as `SRC_LOG_VERBOSE`, `SRC_LOG_QUIET`, and
-`SRC_LOG_SILENT`). Use `logging_settings_from_config()` to build
-`LoggingSettings` from those conventions.
-
-```python
-import src_py_lib as src
+config = src.parse_args(LinearExportConfig, description="Export Linear data")
 
-with src.logging({"src_token": "provided"}):
-    src.info("sync_started", repository_count=3)
-
-    client = src.SourcegraphClient("https://sourcegraph.example.com", "token")
-    data = client.graphql("query Viewer { currentUser { username } }")
+with src.logging(config):
+    client = src.linear_client_from_config(config)
+    client.validate()
+    src.info("Starting export", output_dir=str(config.output_dir))
 ```
 
+- Config precedence is code defaults, `.env`, shell environment, then CLI overrides
+- `parse_args` resolves `op://...` references by default
+- Mark sensitive fields with `secret=True` so config snapshots redact resolved values
+- `src.logging()` configures the root logger by default so project code and `src_py_lib` internals
+  share the same handlers
+- Use `event()` for timed work, `stage()` for workflow context, and
+  `info()` / `warning()` / `error()` for structured one-off events
+
 ## Development
 
 ```sh

pyproject.toml

@@ -11,7 +11,7 @@ dev = [
 [project]
 name = "src-py-lib"
 version = "0.1.0"
-description = "Reusable Python helpers for Sourcegraph projects"
+description = "Reusable libraries for Sourcegraph projects"
 readme = "README.md"
 requires-python = ">=3.11"
 license = "MIT"