This reverts commit 8090d7030984862564a929ee9264e93911014e00.
The cached canonical field was optimizing an unmeasured path: without
the (deferred) O(closure) dependency re-validation multiplier, the
repeated serialization is microseconds for realistic versions. Compute
canonical bytes on demand like the environment, automation, and MCP
stores do, rather than carrying a serde-skipped cache field, a
construction bootstrap, and doubled memory for it. Purely in-memory:
stored blobs and version IDs are unchanged.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Neither method has callers anywhere in the workspace: resolve_reference
splits on '/' directly, and the path-collision validator now checks
ancestor prefixes against a path set. parent() also constructed Self
without going through validate(), so dropping it removes an unvalidated
construction path from the wire type's public API.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The round-trip fixtures only used empty workflow_dependencies, so no
WorkflowVersionId value ever appeared on the wire in a fabro-api
assertion and CreateWorkflowVersionResponse had no coverage at all.
Put a real 64-hex id in the fixture, round-trip the response type, and
pin serialization to the schema's ^[0-9a-f]{64}$ pattern including
lowercase normalization of case-insensitive input.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
WorkflowVersionId bolted a lowercase-only byte scan onto BlobHash
parsing, giving the same 64-hex concept two parse behaviors across
entry points. Identity is the decoded 32-byte digest and canonical
serialization always emits lowercase, so accepting either case on
input is lossless — the stored-blob canonicality check still rejects
non-canonical bytes independently. Delegate straight to BlobHash.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
WorkflowVersion::new serialized the whole version just to enforce the
size limit and threw the bytes away, the store re-serialized them to
write the blob, and every read re-serialized a third time for the
canonicality comparison. Cache the canonical bytes on the struct at
construction (skipped during serde) and expose them as an infallible
borrow; the now-unconstructable InvalidShape store error variant goes
away with it.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
DependencyInvalid fell through to the curated 500 even though the
OpenAPI contract promises 422 workflow_version_dependency_not_found for
an absent, invalid, or non-canonical dependency. Route it to that
response alongside DependencyNotFound; the top-level message only names
the caller-supplied path and id, so no internal chain leaks. Drop the
InvalidVersion/InvalidShape arms, which were unreachable from the only
call site.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The adjacent-pair scan over the byte-sorted path list missed
file/directory collisions whenever a sibling path sorted between the
ancestor and its descendant (any byte below '/' after the shared
prefix, e.g. "assets.txt" between "assets" and "assets/item.txt").
Replace it with an exhaustive ancestor-prefix lookup over a path set,
which also catches equal paths across files and workflow dependencies.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Add the WorkflowVersion domain resource with exactly entrypoint, files,
and workflow_dependencies, plus strict WorkflowPath validation and
deterministic canonical raw JSON. Semantic validation of graph imports,
templates, file references, workflow.toml rules, Dockerfile paths, and
exact child-workflow dependency bindings lives in the new
fabro-workflow-version crate, which validates the complete stored
dependency closure through the shared blob store before writing a root.
The authenticated create-only POST /api/v1/workflow-versions endpoint
ships with its OpenAPI contract, Rust type replacements, and generated
TypeScript client.
Squashed from the resource commits of the original combined branch;
the walker unification this builds on landed separately.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Move the static-reference vocabulary out of fabro-workflow so every
consumer shares one definition: ReferenceKind, AttributeScope, and
reference_kind_for_attribute land in fabro-types::graph, and
validate_static_reference plus a new visit_graph_references walker land
in fabro-template. The manifest bundler drops its ad-hoc graph scan and
walks references through the shared walker.
Unifying the walkers forces three semantic alignments, each matching
what the engine actually executes rather than what the old scanners
happened to match:
- stack.child_dotfile is no longer classified as a child-workflow
reference; the engine never resolved it as one.
- import and stack.child_workflow only count at node scope; graph- and
edge-level occurrences were scanned but never executed.
- @@-escaped goals flow through the shared walker's escape handling
instead of the bundler's own prefix stripping.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Also add environment_images_mut and adopt it in the run compiler's
Dockerfile resolution, replacing the hand-rolled iteration over named
environments plus the run environment.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Route the root workflow through collect_workflow_entry so relative root
arguments are lexically normalized before reading, matching the pre-refactor
behavior: `..` segments no longer resolve through symlinks to a file other
than the one the manifest key names, and `~`-prefixed references are
rejected again. Adds a symlink regression test for the root argument.
Also:
- collect_workflow_entry/collect_workflow_location return the manifest key,
so bundle() no longer recomputes the root key
- hold one FilesystemTemplateStore on the bundler instead of rebuilding it
per template reference
- drop the unused Clone derive on WorkflowScanInput
- replace the hand-rolled JSON literal in the characterization test with an
insta snapshot per the testing strategy
- share one write_file fixture helper between the lib and bundler test
modules
- remove the bundler git-push test; the bundler has no git code path, so the
test could not fail
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Skipping a malformed key under blobs/sha256 during listing now emits a
warn! so operators get a signal when the CAS namespace contains garbage,
matching the projection-cache warmup skip path. Also removes the
runs_share_database_blob_store test, which asserted Arc pointer identity
of internal wiring rather than any observable behavior.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- Collapse RunDatabase::open_writer/open_reader wrappers into one
pub(crate) build, with a Database::open_run_database helper that
gathers the shared-store dependencies in one place
- Stop fetching the blob store on open_run's active-cache hit path
- Share a raw-db test fixture between the two BlobStore raw-key tests
- Evict the cached writer in open_run_reader_is_read_only so the test
exercises the real reader construction path
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Two root-cause fixes for the sandbox failure where an inline Dockerfile
came back from the store as `ARG REDACTED` and the Daytona snapshot
build died on the unset variable.
Entropy redaction measures values, not assignment pairs. The detector
matched `NAME=value` as one token, so an uppercase name merged its
charset into a pure-hex value (which alone can never exceed 4.0 bits)
and pushed the pair over the 4.5-bit threshold — then replaced the
whole pair, destroying the name. `find_entropy_regions` now strips an
identifier-shaped `NAME=` prefix before measuring and redacts only the
value, matching the gitleaks layer's `key=REDACTED` shape.
Execution no longer reads redacted content. Every stored event passes
through the redaction sink, and `load_from_store` rehydrated the
worker's RunSpec from the projection folded from those events — so a
redactor false positive silently rewrote the spec the sandbox builds
from (and changed its snapshot identity). The creation path now writes
the exact spec bytes to the content-addressed blob store and records
`spec_blob` on run.created; `load_from_store` loads the spec from the
blob, keeping the event stream authoritative for run identity,
provenance, and event-recorded blob ids. Retry and fork carry the
source run's `spec_blob` forward, so derived runs stop inheriting the
redacted copy. Runs created before the blob existed fall back to the
folded spec.
The projection and every API surface keep serving the redacted fold;
blobs were already stored unredacted (the workflow bundle carries the
same bytes), so this adds no new exposure at rest.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The entropy redactor rewrites NAME=<hex> assignment pairs to a bare
REDACTED, and the worker rehydrates its executable RunSpec from the
projection folded from redacted stored events. Together these broke
Daytona snapshot builds for any run definition whose inline Dockerfile
pins a git SHA: the spec came back as `ARG REDACTED`, the build died on
the unset variable under `set -eu`, and the environment's snapshot
identity silently changed.
Pin the intended contracts with red tests:
- fabro-redact: an assignment whose value alone is below the entropy
threshold survives redaction (pure hex cannot exceed 4.0 bits; only
the name+value charset merge crosses 4.5), and a genuinely
high-entropy value is redacted without destroying the key name.
- fabro-workflow: the spec that load_from_store rehydrates round-trips
byte-identical through the store, including content that looks like
a secret — event redaction must not reach execution.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The Repo and Workflow dropdowns on the runs page derived their options
from the board query, which is disabled in list view. With ?view=list,
the options were always empty even when runs were visible.
Derive the options from whichever data source the current view loads:
the board query in columns view, or the current page of the paginated
list query in list view. Extract the option-building into an exported
buildFilterOptions helper that also keeps the active selection in the
options when no loaded run matches it, so the filter button never
renders an undefined label while paginating.
A future change will replace page-derived options with a facets
endpoint plus server-side repo/workflow query params.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Run summaries previously populated billing only from the terminal
conclusion event, so the web UI's size chip showed dollar amounts only
after a run completed — even though the size letter was already derived
from live per-stage usage. Derive billing from the same projected total
the size uses. projected_billing already prefers the conclusion's
billing once a run concludes, so completed runs still report the
authoritative final total.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The executor incremented a node's visit count on entry and refused the
visit once the count reached the limit, so a node with max_visits=N
executed at most N-1 times. The documented contract in
stages-and-nodes.mdx is "Max times this node can execute in a run",
and both published examples describe bounded retry loops under that
reading. A graph with max_visits=2 on a designed
one-correction loop therefore failed as "stuck in a cycle" before the
correction could run.
Check the completed-visit count before entry instead: a node with
max_visits=N now executes exactly N times, and the refused entry is
not reported as a visit, so the error's count names the executions
that actually happened. Also correct the nlspec example prose, which
claimed the workflow "moves on with the best result" at the limit;
exceeding max_visits fails the run.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Structural cleanup of the durable pull request creation feature, from a
three-agent review (reuse, quality, efficiency) of the branch:
- Move the supervisor out of handler/ into server/pull_request_supervisor.rs,
collapse its double bookkeeping into one task-id map, and fold the five
copy-pasted failure arms into attempt_pull_request_creation.
- Tag pull_request.failed events with the creation id they resolve, so a
publish-stage failure can never fail an unrelated explicit creation. The
reducer gains PullRequestCreation::succeed/fail transition methods.
- Scan pending creations through a narrow projection-cache accessor instead
of materializing every run summary, raise the scan interval to 30s (notify
covers the live path), and cap retries for runs whose worker cannot even
record a failure.
- Answer "creation already pending" POSTs before taking the per-run create
lock, which a worker can hold for the whole creation.
- Replace the hand-rolled per-run lock map with fabro_store::KeyedMutex.
- Reuse cheap Arc'd projections (cached_run_projection) on the poll endpoint
and in the worker instead of deep-cloning run summaries and diffs.
- Merge ExistingPullRequest into fabro_github::CreatedPullRequest and
extract one reconcile_existing_pull_request helper for both call sites.
- Give the client poll loop a 15-minute deadline; document that Retry-After
and the poll interval are the same constant.
- Resolve a wedged pending creation (run already has a pull request) as a
durable failure instead of skipping it forever.
- Tests: shared wait_for_pull_request_creation helper, a pinned generation-
failure assertion, and a new pipeline test proving reconciliation adopts
an existing PR without an LLM call or create request.
Verified: cargo build --workspace, cargo nextest run --workspace (7,767
passed), nightly clippy -D warnings, fmt --check, insta (no pending), bun
typecheck in fabro-api-client.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Apply cleanup review findings on the collector extraction:
- Deduplicate the lexical path-normalization loop: normalize_absolute_path
now delegates to lexically_normalize_access_path, and it plus
manifest_path_from_absolute live in working_tree.rs so the module
dependency points one way (projection -> collector). Drop the redundant
re-normalization in collect_bundled_file.
- Extract collect_bundled_template_includes to replace the copy-pasted
goal/prompt template-closure sequence, seed_config_document for the
duplicated config seeding, and read_source_input for the duplicated
config reader closures (with the user-settings is_file check hoisted).
- Replace ~100 lines of trivial getters on the Collected* output structs
with pub(super) fields; keep the CollectedPath newtype encapsulated.
- Assemble the manifest by value, moving collected sources into the wire
types instead of deep-copying every file a second time; drop two full
DraftDocument clones that only satisfied the borrow checker; stop
recomputing manifest paths per file in template-dependency verification.
- Resolve the root workflow once in assemble_current_manifest, removing an
unreachable duplicate error path; flatten single-use CollectionNamespace
into a finalize_documents free function.
No behavior change; fabro-manifest tests, clippy, and fmt pass.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Stage the pending task reminder as a Message and add
Message::to_llm_message so durable history and the round-staged turn
share one turn-to-wire conversion. Replace the one-off
BlockingAfterFirstOutputProvider with request capture and an
EventsThenPending variant on ScriptedStreamProvider, add a shared
make_session_with_provider_and_tools helper, and assert the reminder
tests against task_reminder::TASK_REMINDER_TEXT instead of a
substring.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
