mirror of
https://github.com/deepseek-ai/deepseek-harness.git
synced 2026-09-09 04:02:35 +00:00
fix(code-runtime): settle Python provider contracts
This commit is contained in:
@@ -9,7 +9,7 @@ English | [中文](README.zh.md)
|
||||
|
||||
## Summary
|
||||
|
||||
`dsh-experimental-code-runtime-python` ships `PythonCodeRuntime`, the CPython-subprocess implementation of the [`dsh-code-runtime`](../../code-runtime/code-runtime/README.md) seam: it registers as `codeRuntime` with `language: 'python'` and `isolation: 'process'`, spawning a fresh `python3 -I` child per `run()` and executing the program as an async function body over a versionless JSON-lines protocol on the child's fd 3 (stdout/stderr stay free for the program's own output). The host side (`src/protocol.ts`) treats every inbound frame as hostile and rebuilds it before reading; the Python side (`py/protocol.py`) mirrors the message vocabulary. Containment — not a security boundary, model code has bash-equivalent trust — comes from an empty environment, `RLIMIT_CPU`/`RLIMIT_AS`, a wall-clock ceiling, and `SIGTERM`→grace→`SIGKILL` process-group teardown, with all caps validated at plugin load.
|
||||
`dsh-experimental-code-runtime-python` provides the private source-checkout `PythonCodeRuntime`, a CPython-subprocess implementation of the [`dsh-code-runtime`](../../code-runtime/code-runtime/README.md) seam. It registers as `codeRuntime` with `language: 'python'` and `isolation: 'process'`, spawning a fresh CPython 3.10+ child per `run()` and executing the program as an async function body over a versionless JSON-lines protocol on the child's fd 3 (stdout/stderr stay free for the program's own output). The host side (`src/protocol.ts`) treats every inbound frame as hostile and rebuilds it before reading; the Python side (`py/protocol.py`) mirrors the message vocabulary. Containment — not a security boundary, model code has bash-equivalent trust — comes from a tempdir-only environment, `RLIMIT_CPU`/`RLIMIT_AS`, a wall-clock ceiling, and `SIGTERM`→grace→`SIGKILL` process-group teardown, with all caps validated at plugin load.
|
||||
|
||||
## Table of Contents
|
||||
|
||||
@@ -25,11 +25,11 @@ English | [中文](README.zh.md)
|
||||
<a id="use-this-package"></a>
|
||||
## Use this package
|
||||
|
||||
Choose this package to run Python model code through the code-runtime seam: register `PythonCodeRuntime` with `dsh-tools` and `run()` executes each program in a fresh `python3 -I` subprocess, resolving with `result.value` on success and `result.error` on failure (the orthogonal `CodeRunFailure.kind` taxonomy classifies parse failures, thrown exceptions, invalid completions, output overflows, budget expiry, aborts, and substrate death). It rejects only for seam misuse — a malformed binding namespace, or a call after disposal. Configuration is rejected at load: a non-Unix platform, a non-positive or non-integer budget, a `maxLogBytes` below the truncation-marker floor (64), a timer value `setTimeout` would clamp, a budget larger than one fd-3 frame can carry, an `addressSpaceMb`/output-budget pair whose worst-case peak would breach `RLIMIT_AS`, and a `pythonBin` that is not an executable regular file — an explicit path (absolute or containing `/`) is judged directly, a bare name is judged against `PATH`.
|
||||
Choose this private experimental package only in an explicit source-checkout composition. Register `PythonCodeRuntime` beside `dsh-tools` and `run()` executes each program in a fresh CPython 3.10+ subprocess, resolving with `result.value` on success and `result.error` on failure (the orthogonal `CodeRunFailure.kind` taxonomy classifies parse failures, thrown exceptions, invalid completions, output overflows, budget expiry, aborts, and substrate death). It rejects only for seam misuse — a malformed binding namespace, or a call after disposal. Configuration is rejected at load: a non-Unix platform; an explicit `pythonBin` that is not an executable regular file or a bare name that does not resolve on `PATH`; a non-CPython, pre-3.10, or probe-failing interpreter; a non-positive or non-integer budget; a `maxLogBytes` below the truncation-marker floor (64); a timer value `setTimeout` would clamp; a budget larger than one fd-3 frame can carry; or an `addressSpaceMb`/output-budget pair whose worst-case peak would breach `RLIMIT_AS`.
|
||||
|
||||
### What you get
|
||||
|
||||
The package's default export is the `PythonCodeRuntime` plugin. Its public surface also re-exports the host-side protocol vocabulary: `validateChildFrame` (rebuilds every inbound frame), the lossless-JSON codec and meters (`encodeJsonPlain`, `checkDoneValue`, `hasUnsafeIntegerToken`, `hasNonLosslessNumber`), `logTruncationMarker` (the shared truncation-marker text), plus `resolvePythonBin` (interpreter lookup against the current `PATH`), `readProcessStart` (process-start statistics for tests), and `detachResidual` (a test seam for the settled run's resource cleanup). Every cap is a validated `Config` field with a default: `cpuSeconds` (60), `maxWallMs` (600000), `addressSpaceMb` (512, not applied on Darwin), `maxLogBytes` (65536), `maxValueBytes` (32768), `graceMs` (3000), and `pythonBin` (`python3`, resolved before the child spawns with an empty environment: an explicit path must be an executable regular file, a bare name must resolve on `PATH`; either failure is rejected at load, distinguishing 'is not an executable regular file' from 'does not resolve on PATH' instead of silently falling to the platform default `PATH`).
|
||||
The package's default export is the `PythonCodeRuntime` plugin. Its public surface also re-exports the host-side protocol vocabulary: `validateChildFrame` (rebuilds every inbound frame), the lossless-JSON codec and meters (`encodeJsonPlain`, `checkDoneValue`, `hasUnsafeIntegerToken`, `hasNonLosslessNumber`), `logTruncationMarker` (the shared truncation-marker text), plus `resolvePythonBin` (interpreter lookup against the current `PATH`), `readProcessStart` (process-start statistics for tests), and `detachResidual` (a test seam for the settled run's resource cleanup). Every cap is a validated `Config` field with a default: `cpuSeconds` (60), `maxWallMs` (600000), `addressSpaceMb` (512, not applied on Darwin), `maxLogBytes` (65536), `maxValueBytes` (32768), `graceMs` (3000), and `pythonBin` (`python3`, resolved, executable-checked, version-probed, and frozen at load). Each child receives only `TMPDIR`; ambient credentials, `PATH`, `HOME`, and other host state stay unavailable.
|
||||
|
||||
### The wire
|
||||
|
||||
@@ -97,7 +97,7 @@ Read these when the runtime contract is not enough. They move from the seam defi
|
||||
<a id="model-experience"></a>
|
||||
## Model Experience
|
||||
|
||||
Indirectly, through Code Mode in `dsh-tools`, which renders the program's completion value or failure into a retained `run_code` result.
|
||||
Indirectly, through PTC mode in `dsh-tools` when an explicit source-checkout composition mounts this provider; it renders the program's completion value or failure into a retained `run_code` result, and no shipped profile mounts this private package.
|
||||
|
||||
#### KV Cache effect
|
||||
|
||||
@@ -114,7 +114,9 @@ These limits define what the package does and does not cover; they are current p
|
||||
- **A descendant that escapes the child's process group with `setsid()` is not reaped by the group teardown** — `kill(-pid)` cannot reach it; the run still settles on the value the done frame decided, and the close-deadline backstop forces settlement if the orphan holds the pipes open, but the orphan itself outlives the fiber until it exits on its own.
|
||||
- **A `log` frame that arrives after settlement is dropped** — once the run has settled, host-side capture is closed; a late fd-3 `log` frame (from a thread that outlived the done frame) is discarded rather than appended to `logs`.
|
||||
- **A binding REPLY value has no seam-level byte or depth cap** — `maxValueBytes` meters only the done frame's completion value; a wide binding reply is rebuilt host-side (`snapshotJsonValue` traversal) and encoded whole, bounded on both sides only by process memory (like a binding argument, which has no child-side budget either).
|
||||
- **A real-Loader assembly snapshot is deferred to issue #1182 layer 5** — this package is exercised through `ctx.plugin(...)` and real-subprocess tests; the full dsh application composition (codeRuntime registered through a real Loader) is covered by a tracked assembly test in that layer, not by this package's suite.
|
||||
- **No shipped profile mounts this provider** — the keyless `ptc-python-turn` snapshot replaces the headless PTC runtime through the real Loader; released profiles continue to use the worker-thread backend.
|
||||
- **Cross-channel log interleaving is backend-dependent** — Python stdout, stderr, and fd-3 log frames travel independently; each channel preserves its own order, while their total order in `result.logs` may differ.
|
||||
- **CPython 3.10 or newer is required** — the configured executable is resolved and version-probed at load; unsupported interpreters fail before `ctx.codeRuntime` is registered.
|
||||
- **The truncation-marker text and the tempdir prefix keep the pre-rename short names** — the marker `[dsh-code-runtime-python] log capture truncated at <N> bytes` and the `dsh-code-runtime-python-` tempdir prefix are byte-anchored by tests and are independent of the npm package name; promotion (dropping the `experimental-` prefix) does not rename them.
|
||||
- **`run()` is one-shot** — `logs` become available only after `CodeRunResult` resolves; there is no streaming-log or progress interface for output produced by a running program.
|
||||
- **No state persists across runs** — every request executes in a fresh subprocess; a persistent REPL-style kernel stays deferred until a backend brings its own logging scheme.
|
||||
|
||||
Reference in New Issue
Block a user