Documents & Reports — Help Center
Upload SOPs, indexing phases, and report-generation workflows including the auto-generated Feasibility Assessment.
How do I upload documents correctly, and what are the current limits?
Use this workflow to upload documents:
- Open the project chat.
- Use the upload control to add the file.
- Wait for the file status to reach Ready before asking questions against it.
- Ask by topic or by filename when you want precise retrieval.
Supported formats include PDF, DOCX, HTML, spreadsheets, PowerPoint, images, and Markdown.
Current direct-upload behavior:
- The browser uploads large files directly through resumable TUS upload, then PrecisionDocs finalizes and indexes the document.
- Upload progress covers bytes first, then processing: roughly 0-40% for transfer and 40-100% for queueing, parsing, chunking, embedding, and indexing.
- Large PDFs over 5 MB may open a table-of-contents picker so you can ingest only relevant chapters before spending processing time.
- If direct upload is unavailable, the app falls back to the legacy upload path automatically.
Current default limits:
| Limit | Value |
|---|---|
| Max file size | Organization capability limit; current default is 2,000 MB per file |
| Daily upload volume | Organization capability limit; current default is 51,200 MB per day |
| Concurrent processing | Organization capability limit; current default is 16 active files |
| PDF page cap | Organization capability limit; current default is 10,000 pages per PDF |
If a file is rejected, the status indicator will show the configured limit or failure reason so you can trim, split, or choose only relevant chapters.
What do upload statuses mean, and how do cancel or retry work?
Uploaded files move through processing stages before they become searchable.
Common statuses:
- Uploading: the file is being sent to storage
- Queued: accepted and waiting for processing
- Parsing: the platform is extracting text, figures, tables, and page structure
- Chunking: extracted content is being prepared for retrieval
- Embedding: searchable vector and sparse-search entries are being generated
- Indexing: final document records are being written
- Ready: the document is searchable and available to the agent
- Failed or Rejected: the upload did not complete (a human-readable reason is surfaced on the file chip)
Controls:
- Use the X on a processing indicator to cancel.
- Use the retry control on a failed upload to re-run the job.
- For large PDFs, use the TOC picker when it appears if you only need specific chapters.
Do not ask questions against a file until it has reached Ready.
What is the recommended SOP for generating a Due Diligence report?
A Due Diligence report is auto-generated during bootstrap from available API data, so you have a draft immediately on the project home. The agent regenerates it on demand once you have uploaded documents and verified facts in chat.
Best-practice sequence for a high-quality report:
- Confirm bootstrap is complete for both municipality and county.
- Review the auto-generated DD report and Feasibility Assessment to spot obviously wrong facts.
- Upload the critical consultant documents you want referenced (survey, geotech, Phase I ESA, zoning letter).
- Correct any wrong facts in chat — zoning overrides, parcel data corrections, agent directives like "For the stormwater section, lean on the county Title 5 requirements over the city overlay."
- Ask the agent for individual topic checks: flood, wetlands, soils, environmental, ordinance, and major constraints.
- Resolve obvious source conflicts before regenerating the report.
- Regenerate the report only after the factual substrate is in good shape.
Report content reflects the chat:
- Facts you correct in chat win over stale Phase-1 API snapshots when the report is regenerated.
- Per-section directives you give the agent (e.g., "keep the executive summary under 200 words") flow into the writer prompts.
- DD wetland / flood / soil narratives reason over parcel-polygon-clipped acreage (not centroid signals), so you see things like "0.7 acres of Zone AE within parcel boundary" instead of "parcel touches Zone AE".
- The DD report composes the latest Feasibility Assessment section when it exists, so review that deliverable before the final DD export.
Generation expectations:
- Progress is granular from 0-100% across discovery, data collection, extraction, analysis, writing, and QA.
- The QA pass is bounded: it checks for missing required sections, evidence gaps, and obvious contradictions, but it does not replace human engineering review.
- Typical regeneration takes about 1-3 minutes depending on document volume and source complexity.
After generation:
- Review the report against its cited evidence.
- Ask follow-up revision prompts section by section if needed; the agent uses section-level editing instead of rewriting the whole report when you ask for a targeted change.
- Download the DOCX only after the factual sections look right.
How do I use the DocEx editor safely?
The DocEx editor lets you review generated reports, make inline text edits, and export the saved DOCX.
What you can edit:
- Edit existing headings, paragraphs, bullets, callouts, table text, and captions directly on the page.
- Use the right-side contents pane to jump between sections, reorder eligible body sections, or delete an eligible body section.
- Generated front matter is read-only because it refreshes from report metadata during normalization and export.
Structure expectation:
- The editor is text-edit-only; it does not create new sections, tables, figures, or table rows from the document surface.
- Use normal chat report revision prompts when you need the agent to rewrite a section, then reopen or refresh the report version.
- Review tables and figures directly in the editor before exporting.
Autosave and conflicts:
- Unsaved edits are tracked as a dirty state while the editor saves in the background.
- If a newer version is detected while you have local edits, resolve the prompt instead of refreshing the page so you do not lose work.
- Wait for the save state to settle before downloading the final DOCX.
Download expectation: After review, use the report download action to export a DOCX that reflects the saved editor state plus regenerated front matter.
What's in the auto-generated Feasibility Assessment?
A Feasibility Assessment is auto-generated during bootstrap and provides a three-tier risk view immediately, before you upload anything.
Tier 1 — Deterministic site risks (no LLM, pure rules):
- Flood zone classification + acreage within parcel
- Wetlands presence + class + acreage
- Seismic Site Class (ASCE 7-22)
- Elevation profile + slope range (USGS 3DEP)
- Frontage adequacy + ROW classification
- Utility proximity and provider-confirmation caveats when reviewed local utility layers are available
- Annexation or jurisdiction-boundary coordination flags
- PFAS and environmental-screening context where public evidence exists
- Stormwater overlay flags
Tier 2 — Use-case feasibility (LLM-evaluated for the project type):
- Zoning compatibility
- Entitlement path (
by_right/special_use/rezoning_required/variance_required) - Parking / density adequacy
- Environmental + geotech context
Tier 3 — Alternative use scenario comparison — a matrix that scores common alternative project types (e.g., multifamily / industrial / mixed-use) against the same Tier 1 + Tier 2 framework so you can quickly see if a pivot is worth exploring.
The assessment is versioned and downloadable as DOCX. The Feasibility Assessment and DD report remain separate generated deliverables.
Use the assessment as a screening tool. For surveyed elevations, utility availability, wetland delineation, or provider service commitments, upload the authoritative project document or confirm directly with the agency/provider before treating the finding as final.
How do I download reports, maps, and other deliverables?
Deliverables are typically linked directly from the assistant response or available through the project document view.
Typical outputs include:
- Due Diligence reports (DOCX)
- Feasibility Assessments (DOCX, versioned)
- FEMA flood maps (PDF)
- FIRMette / FIRM panel references when FEMA source data supports them
- Wetland maps (PDF)
- Soil reports (PDF)
- Road / frontage maps (PDF)
- Combined site hazard maps (PDF, layered: parcel + flood + wetlands)
- GIS exports (Shapefile .zip, KMZ, KML, GeoJSON, or DXF from the Project Home map)
- Other generated project assets tied to the chat workflow
Download SOP:
- Open the assistant message that generated the asset.
- For generated reports, open the DocEx editor if you need inline text edits, section cleanup, or table/figure review before export.
- Use the download link for the report or map after the save state is current.
- For GIS data, use the Project Home map export dialog and choose the layer set and format for the receiving tool.
- If you need a refreshed version, regenerate or revise the output from chat instead of reusing an outdated file.
Download links are proxied through the app so they remain stable from the user perspective.