Repository navigation
feat(cli): deploy code builds as code bundle uploads - #112
Merged
Merged
Conversation
`pipeline_extract.py --mode code` imports the entrypoint without rewriting it, detects its kind (pipeline, agent, super_component), and returns the platform I/O mapping, a best-effort input JSON schema and a to_dict() graph. Agents and SuperComponents are wrapped like the runtime does (`agent` / `app`), so the mapping comes from the same name-based inference as a YAML build. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Fvxds9E4oHcvivMsoZZpLd
New `_service/code_bundle.py`: finds the project root (nearest pyproject.toml/.git), lists files via git (or a walk) minus `.deepsetignore` and always-excluded secrets/envs, and packs a deterministic tar.gz (sorted, mtime 0, owner 0/"", normalized modes, gzip mtime 0) addressed by sha256, with a 50 MB cap. Dependencies come from [project.dependencies] pinned via uv.lock, else requirements.txt, else the haystack pin. `build_code_bundle` ties it to the extractor's code mode and renders the settings-only config_yaml. Adds tomli for Python 3.10. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Fvxds9E4oHcvivMsoZZpLd
- DeploymentsAPI.get_code_bundle / upload_code_bundle against /code_bundles, mapping the typed error contract: typed 404 -> upload, typed 403 -> "not enabled for this organization", untyped 404 -> "use --build yaml". push_revision takes an optional code_bundle field. - DeploymentService.deploy(build="code") packs the project, uploads the bundle unless the server has it, and only then resolves/creates the service and pushes the settings-only config_yaml with code_bundle. Platform YAML validation is skipped for code builds. - `deploy --build auto|yaml` (auto = yaml). `code` is accepted but undocumented for now; it takes a directory target with --entrypoint path.py:attr, and --dry-run --output writes the tar.gz. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Fvxds9E4oHcvivMsoZZpLd
hub-api renders a dict detail as the whole body ({code, message}); the
nested {detail: {code, message}} shape is still accepted.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Fvxds9E4oHcvivMsoZZpLd
A SuperComponent spread over two modules with no LLM or secrets, to try code builds end to end. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Fvxds9E4oHcvivMsoZZpLd
A deepset.yaml in the project root names the entrypoint, kind and dependencies of a code build; with one present, `deploy --build auto` builds as code. A plain callable entrypoint is a `function` build: fixed payload/result mapping, input schema from its signature, and deploy prints the /run endpoint instead of chat completions. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Fvxds9E4oHcvivMsoZZpLd
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Fvxds9E4oHcvivMsoZZpLd
`invoke <service> --inputs JSON` calls a function code build's /run, `--query TEXT` goes through OpenAI-compatible chat completions for pipeline, agent and SuperComponent services. The platform picks the error for the wrong route, so the CLI shows its message as-is. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vmfn8GHFL2DrPGk7XjbQvN
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vmfn8GHFL2DrPGk7XjbQvN
`Path('.')` has no name, so the io-config lookup crashed and the revision
comment and source metadata came out empty. Resolve the target first.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Vmfn8GHFL2DrPGk7XjbQvN
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vmfn8GHFL2DrPGk7XjbQvN
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Vmfn8GHFL2DrPGk7XjbQvN
Only a code build can take a directory, so --build auto no longer needs a deepset.yaml to pick it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The entrypoint comes from --entrypoint, dependencies from the io-config or pyproject.toml, and the kind is always detected. auto builds a directory target as code. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
CI forces a color terminal, and the escape codes split "--build". Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
mrchtr
marked this pull request as ready for review
October 28, 2026 12:40
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0154qh65U5VAn77QjCc5LH99
Contributor
Coverage report
Click to see where and how coverage changed
This report was generated by python-coverage-comment-action |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
List `code` in `deploy --build` help as experimental instead of hiding it, and warn on stderr whenever a deploy resolves to a code build (also via `auto` on a directory), since hiding the value never stopped `auto` from reaching it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0154qh65U5VAn77QjCc5LH99
volkanarisli
approved these changes
Oct 30, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Adds code builds to the CLI: instead of inlining a pipeline into YAML, a project is packed into a code bundle and uploaded to the platform.
_service/code_bundle.py: packs a project into a deterministic code bundlepipeline_extract: extractor code mode for code builds_api/deployments.py: code bundle upload endpoints, reading hub-api's flat typed error bodydeployis gated behind a platform capability check.) is treated as a code build, with the entrypoint given by--entrypoint path.py:attr