Upload and Ingestion
Upload PDFs and monitor the ingestion pipeline as documents are parsed, chunked, and indexed.
What this page does
The Upload page is where you add new documents to Planara's knowledge base. It has two sections: the upload form and the ingestion jobs list.
Uploading a document
Step 1: Select a file
Drag and drop a PDF onto the drop zone, or click to browse. Only PDF files are accepted. After selecting a file, you will see:
- File name and size
- A "Remove" button to clear the selection
Step 2: Set document metadata
| Field | Description |
|---|---|
| Document Name | Display name for the document. Auto-populated from the filename but editable. Example: "Yamaha F350A Service Manual" |
| Document Type | Dropdown to classify the document. This affects how Planara processes and retrieves it. |
| Equipment Models | Tag the document with the equipment models it applies to. This scopes retrieval so technicians only see documentation for the equipment they are working on. Use "Global" for documents that apply across all models (general safety procedures, shop standards). |
Document types
| Type | Label | Description |
|---|---|---|
owners_manual | Owner's Manual | End-user documentation with simplified procedures |
service_manual | Service Manual | Technical documentation for trained technicians |
parts_catalog | Parts Catalog | Part numbers, exploded diagrams, ordering information |
tsb | Technical Service Bulletin | Manufacturer-issued fixes and updates |
ckl | Safety / Compliance Standard | Ingested into the CKL index, not the main search index |
training_guide | Training Guide | Training materials and educational content |
custom | Other / Custom | Documents that do not fit other categories |
Important: Selecting "CKL" routes the document to the Compliance Knowledge Layer index. It will be used for safety enrichment, not primary retrieval. See CKL Explained for details.
Step 3: Upload
Click "Upload & Ingest" to:
- Upload the PDF to secure cloud storage
- Create an ingestion job record
- Queue the document for processing
A green confirmation banner appears: "Document uploaded and queued."
Ingestion pipeline
After upload, the document goes through this automated pipeline. Each stage is shown as a progress step in the job card:
- Queued -- waiting for processing capacity. The ingestion worker picks up jobs automatically.
- Processing -- the document is being read page by page:
- Multiple AI parsers extract content (multi-parser extraction)
- Content is classified into semantic types (specs, procedures, safety warnings, diagrams)
- Entities are extracted (part numbers, components, specs, tools)
- Chunks are created preserving semantic units
- Tables are indexed with triple representation (structured, natural language, key-value)
- Multi-page procedures are auto-linked
- High-dimensional semantic representations are generated
- Content is stored in the vector index with hybrid keyword + semantic indexing
- QA gates -- 7 automated quality checks run before chunks are committed. If any gate fails, the job pauses with a failure reason.
- Complete -- the document is fully indexed and searchable
- Failed -- something went wrong (error message shown)
Ingestion jobs list
Below the upload form, all ingestion jobs for your namespace are listed newest-first.
Job information
Each job card shows:
| Field | Description |
|---|---|
| Document name | The name you assigned during upload |
| Version badge | Shows v2, v3 etc. if this document has been re-uploaded |
| CKL badge | Blue badge if this is a CKL document |
| Status badge | Queued, Processing, Complete, Failed, Paused, or Cancelled |
| Timestamps | Created, Started, Completed (or Elapsed time for in-progress jobs) |
| Supersedes | If this is a re-upload, shows which previous job it replaces |
| Progress bar | For processing jobs: pages processed out of total pages, with percentage |
| Error message | For failed jobs: the specific error that occurred |
Job controls
| Button | Description |
|---|---|
| Pause | Suspends a processing job. Chunks already processed are retained. |
| Resume | Resumes a paused job from where it left off. |
| Cancel | Requests cancellation. Shows "Cancellation requested..." until the system confirms. |
Auto-refresh
When any job is in "processing" or "queued" state, the page auto-refreshes every 10 seconds to show progress updates. A blue pulsing indicator in the header confirms auto-refresh is active.
Re-uploading a document
If you upload a document with the same name as an existing one, Planara auto-increments the version number. The new version replaces the old one in the search index once processing completes. The "Supersedes" field on the job card shows which previous job is being replaced.
FAQ
Q: How long does ingestion take? A: Depends on document size. A 100-page service manual typically takes 5-10 minutes. Large documents (500+ pages) may take 30+ minutes.
Q: Can I upload non-PDF files? A: Not currently. Only PDF files are accepted.
Q: What happens if ingestion fails? A: The job card shows the error message. Common causes: corrupted PDF, API timeout, storage quota exceeded. You can re-upload the document to try again.
Q: Does re-ingestion affect existing answers? A: Yes. When a document is re-ingested, old chunks are replaced with new ones. Answers to queries that relied on the old chunks may change. Run an eval after re-ingestion to verify.