René's URL Explorer Experiment

Title: feat: stream CLI log transcripts and run status to Socket backend by barslev · Pull Request #201 · SocketDev/socket-python-cli · GitHub

Open Graph Title: feat: stream CLI log transcripts and run status to Socket backend by barslev · Pull Request #201 · SocketDev/socket-python-cli

X Title: feat: stream CLI log transcripts and run status to Socket backend by barslev · Pull Request #201 · SocketDev/socket-python-cli

Description: Adds an opt-in streaming log channel between the Python CLI and the Socket backend so the CLI run's terminal status (in_progress / success / failure / cancelled) and a transcript of its own log output are visible in the Socket admin views when the user opts in with --upload-logs. The Socket backend may also force-enable streaming for specific orgs regardless of the flag. Why? The Socket backend currently has no visibility into what happens inside a CLI invocation — there's no record of whether a scan ran to completion or what was logged along the way. This PR opens a bounded, opt-in side-channel that uploads the CLI's own log records to the backend and reports the run's terminal status, without changing any existing CLI request on the wire. Changes New modules socketsecurity/core/cli_run.py — register_cli_run and finalize_cli_run lifecycle helpers, both best-effort. register_cli_run always calls POST /v0/python-cli-runs with the user's share_logs choice; the server responds with log_streaming_enabled and a nullable run_id. The CLI gates the rest of the lifecycle on log_streaming_enabled (the server can force-enable for an org even when the client didn't ask). finalize_cli_run accepts an optional report_run_id that links the run to the full-scan it produced. socketsecurity/core/log_uploader.py — BatchedLogUploader (daemon-thread flusher, 5s flush interval, swallows all network errors at-most-once, skips empty buffers, drains on shutdown) plus UploadingLogHandler — a logging.Handler that routes records to the uploader. Includes a thread-local recursion guard so the uploader's own request-error logs aren't re-enqueued mid-flush. socketsecurity/core/streaming.py — setup_streaming() wires the uploader and a terminal handler into the socketcli and socketdev loggers, forces both loggers to DEBUG so the upload captures the full history regardless of --enable-debug state, and returns a teardown callable for the caller to register with atexit. Module-level setters set_run_status() and set_report_run_id() propagate the terminal status and the associated full-scan id into finalize_cli_run. Existing files socketsecurity/config.py — adds --upload-logs / --upload_logs opt-in flag (advanced group, default off). When set, the CLI sends share_logs=true in the register payload. socketsecurity/socketcli.py — always calls setup_streaming() after CliClient init (register is cheap and the server may force-enable for an org); registers the teardown via atexit only when the server returned a run_id. Captures diff.id at a single chokepoint after the diff-producing branches converge and threads it through set_report_run_id(), guarded against the NO_DIFF_RAN / NO_SCAN_RAN sentinel values. Exception handlers in cli() now call set_run_status(...): KeyboardInterrupt → cancelled any SystemExit with non-zero code → failure uncaught Exception → failure clean exit → default success The SystemExit handling is load-bearing: several failure paths in main_code() call sys.exit(3) directly, which bypasses except Exception (since SystemExit is a BaseException subclass, not Exception). Test plan uv run pytest tests/unit — 171 unit tests pass, including the new tests in test_cli_run.py, test_log_uploader.py, and test_streaming.py (covering opt-in/opt-out register paths and the disabled-by-server response shape). End-to-end against a local Socket backend with the matching endpoints applied: --upload-logs set, server says enabled → full CLI log captured, run reports status: success, report_run_id populated with the full-scan id and joinable to the underlying full_scans row --upload-logs set, simulated RuntimeError → run reports status: failure, log captured up to the crash 15s continuous-log run → batches arrive at the backend at ~5s intervals while the CLI is still running Without --upload-logs (and no org-side override) → CLI calls register, gets log_streaming_enabled: false, sends no further streaming requests; scan otherwise unchanged Register failure (404 / 401 / network down) → CLI silently degrades, scan continues normally, no exception surfaces Public Changelog A new opt-in --upload-logs flag uploads the CLI's log output to the Socket backend for the duration of the run, alongside a per-run status (in_progress / success / failure / cancelled). The transcript is captured regardless of the local --enable-debug state; the existing terminal verbosity is unchanged. The Socket backend can also force-enable streaming for specific orgs regardless of the flag. Default off; CLI runs without the flag are unaffected.

Open Graph Description: Adds an opt-in streaming log channel between the Python CLI and the Socket backend so the CLI run's terminal status (in_progress / success / failure / cancelled) and a transcript of its own log...

X Description: Adds an opt-in streaming log channel between the Python CLI and the Socket backend so the CLI run's terminal status (in_progress / success / failure / cancelled) and a transcript of its own...

Opengraph URL: https://github.com/SocketDev/socket-python-cli/pull/201

X: @github

direct link

Domain: github.com

Links:

Viewport: width=device-width