Jobber/orchestrator.md at 7f517776dfb49338f0b4ad0c9ce0600204c5927a

Shaheer Sarfaraz f3c164d252

feat(pipeline): parallelize discovery/process via evolved asyncPool (#211 )

* feat(pipeline): centralize concurrency hooks and parallelize discovery/process steps

* feat(orchestrator): unify single and bulk job actions API

* job actions de-bulk-ified

* application inbox section debulk

* chore(orchestrator): remove remaining bulk wording from job action flow

* select multiple to skip with shortcut

* comments

* coomeents

* fix progress ordinal and add jobs actions payload examples

2026-02-20 16:49:13 +00:00

3.9 KiB

Raw Blame History

id, title, description, sidebar_position

id	title	description	sidebar_position
orchestrator	Orchestrator	Job states, ready flow, and PDF generation/regeneration behavior.	1

What it is

The Orchestrator is the primary jobs workspace in JobOps.

It controls:

job lifecycle states
manual and automatic ready flow
PDF generation and regeneration
handoff to post-application tracking

Job states:

discovered: found by crawler/import, not tailored yet
processing: tailoring and/or PDF generation in progress
ready: tailored PDF generated and ready to apply
applied: marked as applied
skipped: explicitly excluded from active queue
expired: deadline passed

Why it exists

Orchestrator centralizes the transition from discovered opportunities to application-ready artifacts.

It exists to ensure:

a consistent path from discovery to tailored output
clear status transitions across manual and automated workflows
predictable regeneration behavior when job data changes

How to use it

Intended ready flow

Manual flow:
1. Job starts in discovered.
2. Open the job and choose Tailor.
3. Edit JD/tailored fields/project picks.
4. Click Finalize & Move to Ready.
Auto flow:
1. Pipeline scores discovered jobs.
2. Top jobs above threshold are auto-processed.
3. Jobs move directly to ready with generated PDFs.

Ghostwriter availability

Ghostwriter is available in discovered and ready job views.

For details, see Ghostwriter.

Opening documentation from the sidebar

Open the sidebar menu.
In the footer section under Version <build>, click Documentation, which opens the locally hosted docs in a new tab.

Generating PDFs

PDF generation uses:

base resume selected from RxResume
job description
tailored summary/headline/skills/projects

Common paths:

Discovered to finalization: POST /api/jobs/actions with payload:

curl -X POST "http://localhost:3001/api/jobs/actions" \
  -H "content-type: application/json" \
  -d '{
    "action": "move_to_ready",
    "jobIds": ["<jobId>"]
  }'

Streaming progress: POST /api/jobs/actions/stream (same JSON payload)
Ready regeneration: POST /api/jobs/:id/generate-pdf

Regenerating PDFs after edits (copy-pasteable examples)

If JD or tailoring changes, regenerate PDF to keep output in sync.

curl -X PATCH "http://localhost:3001/api/jobs/<jobId>" \
  -H "content-type: application/json" \
  -d '{
    "jobDescription": "<new JD>",
    "tailoredSummary": "<optional>",
    "tailoredHeadline": "<optional>",
    "tailoredSkills": [{"name":"Backend","keywords":["TypeScript","Node.js"]}],
    "selectedProjectIds": "p1,p2"
  }'

curl -X POST "http://localhost:3001/api/jobs/<jobId>/summarize?force=true"
curl -X POST "http://localhost:3001/api/jobs/<jobId>/generate-pdf"

External payload and sanitization defaults

LLM prompts send minimized profile/job fields.
Webhooks are sanitized and whitelisted by default.
Logs and error details are redacted/truncated by default.
Correlation fields include requestId, and when available pipelineRunId and jobId.

Common problems

Job is stuck in `processing`

processing is transient; failures generally revert the job to discovered.
Check run logs and retry generation.

PDF does not reflect recent edits

Run summarize with force=true after changing the JD/tailoring.
Regenerate PDF after summarize completes.

Reopen skipped/applied jobs

Patch status back to discovered to return the job to the active queue.

3.9 KiB Raw Blame History