Financial Report Parser
The Financial Report Parser template extracts structured financial metrics from unstructured report text. Given a financial report (earnings release, 10-Q, analyst summary), it identifies key figures — gross profit, EBITDA, net income, operating expenses — and returns them as structured, consistently formatted data.
How the flow works
_10Chat Input → Language Model → Structured Output → Parser → Chat Output
The Chat Input component receives the financial report text (a sample is pre-loaded in the template). The Language Model analyzes the text and extracts the target metrics. The Structured Output component enforces a consistent schema on the extracted data. The Parser formats the result into readable text, which appears in Chat Output.
Prerequisites
- An API key for a supported LLM provider (e.g., OpenAI)
Quickstart
-
In Orbis Studio, open the Templates library and select Financial Report Parser.
-
In the Language Model component, add your LLM API key. A sample financial report is pre-loaded in Chat Input — you can run the flow immediately without supplying your own data.
-
Open the Playground and click Run (or press Enter with the default input). The flow extracts and displays the financial metrics.
-
To parse your own report, replace the pre-loaded text in Chat Input with your document content.
Key components
| Component | Role |
|---|---|
| Chat Input | Receives the financial report text; pre-loaded with a sample report |
| Language Model | Reads the report and extracts the target financial metrics |
| Structured Output | Enforces a consistent output schema (gross profit, EBITDA, net income, operating expenses) |
| Parser | Formats the structured data as human-readable text |
| Chat Output | Displays the extracted metrics in the Playground |
Configuration tips
- Custom metrics: Edit the Structured Output schema to extract different or additional fields — for example, revenue, EPS, or segment-level breakdowns.
- Batch processing: To process multiple reports, connect a Loop component upstream to feed reports one at a time through the same pipeline.
- Output format: Modify the Parser template to change how results are displayed — for example, as a table, JSON, or a narrative summary.
- Model selection: Larger models (GPT-4o, Claude 3.5 Sonnet) tend to be more accurate on complex financial language. Test with a smaller model first for cost, then upgrade if extraction quality is insufficient.
Next steps
- Store extracted metrics to a database or spreadsheet using an API Request component or external integration.
- Add a Knowledge Ingestion step to build a searchable archive of parsed reports over time.