CSV Upload + Go Fig
Upload CSV files directly to Go Fig for quick data integration without technical setup.
Not every source of finance data has an API. Commission schedules, acquisition-era legacy exports, manual operational trackers, board-approved budgets, and bank-reconciliation files often live as CSVs. Go Fig's CSV upload is a first-class path into the Financial Intelligence Graph, with the governance that custom data usually lacks. Files can be dropped through the UI, dropped into a watched S3-compatible bucket, or posted through a signed upload URL from a workflow. Go Fig infers types from the first 1000 rows, flags encoding issues (UTF-8, Latin-1, CP1252), and tracks schema between versions so a renamed column or a new row of totals at the bottom does not silently break downstream analyses. Every upload is versioned, every schema change is logged, and every row traces back to the file and uploader that produced it. The goal is the governance of a piped connector with the flexibility of a spreadsheet handoff.
Key facts
- Schema inference
- First 1000 rows
- Versioning
- Every upload snapshotted
- Drop modes
- UI, S3-compatible bucket, signed URL
- Encoding
- UTF-8, Latin-1, CP1252 auto-detected
- File size
- Up to 2 GB per upload
What you can do with CSV Upload data in Go Fig
Governed manual handoffs
Replace email-attached spreadsheets with versioned, schema-tracked uploads that Celeste can query alongside connected systems.
Legacy system bridge
Ingest exports from on-premise ERPs, acquired-company systems, or any tool without an API, without losing the audit trail.
Budget and forecast ingest
Drop the approved budget file and join it automatically to GL actuals for variance analysis, no modeling tool required.
Data available from CSV Upload
Go Fig extracts and normalizes the following data from your CSV Upload account:
How to connect CSV Upload
Pick a drop method
For ad-hoc uploads, use the web UI (drag and drop, authenticated). For automated handoffs, provision an S3-compatible bucket prefix and use the issued credentials inside your extract script. For signed-URL posting, Go Fig issues short-lived URLs from the API that your workflow tool can POST to.
Declare the schema on first upload
Go Fig infers column types from the first 1000 rows but prompts you to confirm. Declaring the schema means subsequent uploads are validated against it, so a column-rename or an inserted total-row at the bottom of the file is caught at ingest instead of breaking a dashboard silently.
Set the upload cadence and retention
Uploads can be one-shot (a backfill) or recurring (monthly commission file). For recurring uploads, configure the expected cadence and retention. Missed uploads surface as a connector-health alert in the admin view.
Join to connected data in the graph
Once a file is governed, it is queryable like any connected source. Starter patterns exist for budget-vs-actual, commission-to-GL, and acquired-company GL bridging. Celeste can query it in natural language and the uploader and file version are tracked in every answer's audit trail.
Authentication: Direct file upload via authenticated web UI, or programmatic upload via S3-compatible bucket with per-user IAM credentials. Watched-bucket mode uses a bucket prefix and role assumption so automated extracts from legacy systems can drop files without a human in the loop.
Common Questions About CSV Upload Integration
How does Go Fig infer column types from a CSV?
Go Fig samples the first 1000 rows and proposes a type per column (string, integer, decimal, date, datetime, boolean). Ambiguous columns (all nulls, or values that could parse as multiple types) prompt for explicit user confirmation. The declared schema is locked on the first upload so subsequent uploads are validated, not reinferred, which prevents silent type drift between monthly files.
What happens when a CSV schema changes between uploads?
A new column, a renamed column, or a dropped column triggers a schema-change event. The upload is held in a staging state and the admin is notified. Nothing downstream breaks silently. For expected changes (a new cost-center column added this year), you approve the change and the schema updates with versioning. For unexpected changes, you reject and the uploader fixes the export.
Can uploads be automated from legacy systems without an API?
Yes. The S3-compatible bucket drop mode is the most common pattern: a cron job on the legacy system exports a file and writes it to the bucket, Go Fig picks it up within a minute and ingests it. For finance teams that have standardized on Excel-based workflows, a watched SharePoint or Google Drive folder also works, using the corresponding connector as the intermediary.
How are CSV uploads governed for audit?
Every upload carries metadata: who uploaded it, when, what file version, what schema version, and what rows changed versus the prior version. When Celeste answers a question that touches CSV data, the audit trail shows exactly which file version the answer came from. This is the same audit pattern used for API-sourced data, just with a human upload event instead of a polling timestamp.
What file sizes and formats are supported?
Up to 2 GB per upload, with larger files handled via multipart upload in the automated modes. CSV, TSV, pipe-delimited, and Excel-exported variants are all supported, as are .gz and .zip compressed archives. Encoding is auto-detected (UTF-8, Latin-1, CP1252) with explicit override available for edge cases like files produced by older Windows tools.
Ready to connect CSV Upload?
See how your CSV Upload data looks in Go Fig with a personalized demo.
Book a Demo