Pipeline Result

[MODE:DIALOG] Read-only, idempotent. Tier: meta.

Looks up the cached output of a prior fibe_pipeline invocation. Cache is per-session, TTL 5 minutes. Both successful and partial-failure results are cached.

When to use

• Pipeline returned a large bindings map and you want a single field without re-running.
• Pipeline failed mid-run; need to inspect specific step outputs to plan cleanup.
• Resuming work after switching task and losing the original tool result from your context.

Inputs

Field	Type	Required	Notes
pipeline_id	string	yes	From the prior pipeline's pipeline_id
path	string	no	JSONPath; rooted at step bindings, falls back to full response

Path resolution

• "$.step_id.field": looks up inside bindings (the steps map).
• "$.status", "$.error", "$.completed_step_ids": fall back to the full cached envelope.
• Empty path returns the entire cached response.

Output

On hit: the projected JSON or full cached object.

On expiry/miss:

{ "expired": true, "pipeline_id": "<id>" }

Gotchas

• Cache is keyed by session ID, not pipeline ID alone: another session cannot read your cache.
• TTL is 5 minutes hardcoded; long-running flows that wait then look up should re-run if past that window.
• path is JSONPath (PaesslerAG/jsonpath v1): slice indexing, recursive descent, and filter expressions all work.

Recipe

After a fibe_pipeline returns pipeline_id="abc":

• Get the failed step's API code: {pipeline_id:"abc", path:"$.error.code"}.
• Get just one created ID: {pipeline_id:"abc", path:"$.create_pg.id"}.

Related

• fibe_pipeline: the producer.