> ## Documentation Index
> Fetch the complete documentation index at: https://docs.fastino.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Example use cases

Use GLiNER when your workflow needs fast, structured decisions from text. Use Fastino-Nemotron-3.5-Lightning-Healthcare or Fastino-Nemotron-3.5-Lightning-Finance when it needs domain-specific language understanding and generation.

## Choose a model

| Need                                                         | Start with                                |
| ------------------------------------------------------------ | ----------------------------------------- |
| Extract entities, labels, relationships, or policy signals   | GLiNER                                    |
| Route an agent or decide whether a person should review work | GLiNER                                    |
| Summarize, reason over, or draft healthcare content          | Fastino-Nemotron-3.5-Lightning-Healthcare |
| Analyze financial documents, tables, and calculations        | Fastino-Nemotron-3.5-Lightning-Finance    |

## GLiNER: structured decisions for AI systems

### Agent and browser workflows

* **Browser use and automation:** identify page intent, relevant fields, warnings, and actions that need confirmation.
* **Context compaction:** extract decisions, tasks, deadlines, and user preferences from conversations or agent traces.
* **Model routing:** classify domain, complexity, latency requirements, and risk before selecting a model or workflow.
* **Agent supervision:** detect drift, repeated failures, missing deliverables, risky tool calls, irrelevant context, poor output quality, or when a human should step in.

### Retrieval, data, and research

* **RAG:** select relevant passages before generation and check retrieved context for relevance or conflicts.
* **Knowledge graphs and agent memory:** extract entities and relationships from documents or conversations into a queryable fact layer.
* **Text classification:** sort emails, tickets, and documents into operational labels such as spam, urgent, refund request, or sales inquiry.
* **PII detection and redaction:** identify sensitive data before storage, indexing, or model calls.
* **Forecasting and market intelligence:** turn customer feedback, filings, transcripts, and news into signals such as demand, supply concerns, competitor mentions, and management changes.

### Safety and quality checks

* **LLM guardrails and moderation:** screen prompts and outputs for harmful content, sensitive data, and policy violations.
* **Prompt-injection detection:** flag attempts to override instructions, exfiltrate data, or manipulate tool use.
* **Semantic code linting:** check generated code and pull requests for conventions that syntax alone cannot catch.

<Warning>
  For agent monitoring, evaluate only artifacts your system is allowed to collect, such as tool calls and user-visible logs. Do not use hidden chain-of-thought as an operational monitoring surface.
</Warning>

## Fastino-Nemotron-3.5-Lightning-Healthcare

* **Clinical documentation:** summarize notes, encounters, patient timelines, and discharge documentation.
* **Chart preparation:** surface relevant diagnoses, medications, symptoms, and recent events before an encounter.
* **Patient communication:** draft plain-language, clinician-approved care instructions and responses.
* **Clinical research:** organize evidence, summarize literature, and support trial-matching review.
* **Documentation review:** flag missing sections, conflicting dates, and unclear follow-up items.

<Warning>
  This model supports clinical workflows. Qualified clinicians should review outputs before they affect patient care.
</Warning>

## Fastino-Nemotron-3.5-Lightning-Finance

* **Filing and annual-report Q\&A:** answer research questions over reports, filings, and dense tables.
* **Financial document summaries:** produce structured notes from earnings releases, investor presentations, and reports.
* **Table and metric analysis:** locate relevant figures, calculate comparisons, and show the formula used.
* **Multi-turn financial research:** continue a line of inquiry while keeping the relevant sources and assumptions visible.
* **Earnings review and forecasting inputs:** extract reported metrics, guidance, narrative signals, and market changes for analyst or forecasting workflows.
* **Compliance-aware research:** pair generation with GLiNER checks for PII, required disclosures, sensitive claims, and review routing.

<Warning>
  This model supports financial research and operations. Validate calculations, sources, and material decisions with qualified professionals.
</Warning>

## Combine models in one workflow

| Workflow                     | GLiNER                                               | Lightning model                                 |
| ---------------------------- | ---------------------------------------------------- | ----------------------------------------------- |
| Clinical message routing     | Detect intent, urgency, PII, and escalation triggers | Draft a response from approved clinical context |
| Financial research assistant | Select relevant filings and policy checks            | Answer with citations and calculation steps     |
| Controlled agent             | Score tool-call risk and completion                  | Handle the specialized language task            |

Start with a measurable task and a small evaluation set built from permissioned examples. For GLiNER, define the labels or schema. For a Lightning model, define the source documents, required checks, and review step.