Ollama connectors

Integrate local LLMs via Ollama into Bonita processes for text generation, classification and data extraction without sending data to external APIs.

The Ollama connector is part of the Bonita AI Connectors family.

Getting started

Import the bonita-connector-ai-ollama module as an extension dependency in your Bonita project. See the AI connectors overview for general setup instructions.

You must also have Ollama installed and running, either on the same server as Bonita or on an accessible network host.

Connection configuration

Parameter Required Description Default

Parameter	Required	Description	Default
API Key	No	Not required for Ollama (use any non-empty value or leave blank)
Base URL	No	Ollama server endpoint	`http://localhost:11434`
Model Name	No	Local model to use	`llama3.1`
Temperature	No	Controls randomness (0.0 to 1.0)
Timeout	No	Request timeout in milliseconds

API Key

Not required for Ollama (use any non-empty value or leave blank)

Base URL

Ollama server endpoint

http://localhost:11434

Model Name

Local model to use

llama3.1

Temperature

Controls randomness (0.0 to 1.0)

Timeout

Request timeout in milliseconds

Ollama runs entirely on-premises, making it suitable for sensitive data that cannot leave the organization’s network.

Available models

Ollama supports any model from the Ollama Library. Common models include:

llama3.2 — Meta’s latest, good general-purpose model
llama3.1 (default) — Previous generation, well tested
mistral — Mistral 7B, fast and efficient
codellama — Optimized for code-related tasks
gemma2 — Google’s open model
phi3 — Microsoft’s small but capable model
llama3.2:1b — Very small, fastest inference

To install a model, run:

ollama pull llama3.2

Operations

Ask

Send a user prompt (with optional system prompt and documents) to a local model and return the generated response.

Parameter Required Description Default

Parameter	Required	Description	Default
User Prompt	Yes	The prompt to send to the model
System Prompt	No	System instructions to guide the model behavior	`You are a polite assistant.`
Output JSON Schema	No	JSON Schema to structure the response as JSON
Source Document Reference	No	Bonita process document to include as context
Source Document References	No	List of Bonita process documents to include as context

User Prompt

Yes

The prompt to send to the model

System Prompt

System instructions to guide the model behavior

You are a polite assistant.

Output JSON Schema

JSON Schema to structure the response as JSON

Source Document Reference

Bonita process document to include as context

Source Document References

List of Bonita process documents to include as context

Parameter	Type	Description
output	String	The generated response from the model

Parameter

Type

Description

output

String

The generated response from the model

Classify

Classify a document into one of the predefined categories.

Parameter	Required	Description
Categories	Yes	Comma-separated list of classification categories
Source Document Reference	Yes	Bonita process document to classify
Source Document References	No	List of documents to classify

Parameter

Required

Description

Default

Categories

Yes

Comma-separated list of classification categories

Source Document Reference

Yes

Bonita process document to classify

Source Document References

List of documents to classify

Parameter Type Description

Parameter	Type	Description
output	String	JSON with `category` and `confidence` fields

output

String

JSON with category and confidence fields

Sample classification result

{
  "category": "INTERNAL_MEMO",
  "confidence": 0.82
}

Extract

Extract structured data from a document using field names or a JSON Schema.

Parameter	Required	Description
Fields to Extract	No	Comma-separated list of field names to extract
Output JSON Schema	No	JSON Schema defining the extraction structure
Source Document Reference	Yes	Bonita process document to extract from
Source Document References	No	List of documents to extract from

Parameter

Required

Description

Default

Fields to Extract

Comma-separated list of field names to extract

Output JSON Schema

JSON Schema defining the extraction structure

Source Document Reference

Yes

Bonita process document to extract from

Source Document References

List of documents to extract from

You must provide at least one of fieldsToExtract or outputJsonSchema parameters.

Parameter	Type	Description
output	String	JSON with extracted fields

Parameter

Type

Description

output

String

JSON with extracted fields

Use cases

Development and testing

Use Ollama during development and testing phases to iterate on AI prompts and configurations without incurring API costs or sending sensitive test data to external services.

Process flow:

Developer configures a connector in Bonita Studio with Ollama
Prompts and JSON schemas are tested locally with instant feedback
Once validated, the connector configuration is switched to a cloud provider for production
The same prompt and schema work across all providers

Configuration:

{
  "apiKey": "not-needed",
  "url": "http://localhost:11434",
  "chatModelName": "llama3.2",
  "systemPrompt": "You are a document analysis assistant. Return structured JSON only.",
  "userPrompt": "Extract the key dates, amounts, and parties from this invoice.",
  "outputJsonSchema": "{\"type\":\"object\",\"required\":[\"invoiceNumber\",\"issuer\",\"totalAmount\",\"dueDate\"],\"properties\":{\"invoiceNumber\":{\"type\":\"string\"},\"issuer\":{\"type\":\"string\"},\"totalAmount\":{\"type\":\"number\"},\"dueDate\":{\"type\":\"string\"}}}"
}

Expected output:

{
  "invoiceNumber": "INV-2026-TEST-001",
  "issuer": "Test Company Ltd",
  "totalAmount": 500.00,
  "dueDate": "2026-04-30"
}

Air-gapped environments

Run AI-powered document processing in environments with no internet access, such as government, defense, or highly regulated industries.

Process flow:

Documents are uploaded to the Bonita process in the air-gapped network
A service task uses the Extract connector pointed at the local Ollama server
All processing happens within the network perimeter
No data leaves the secured environment

Configuration:

{
  "apiKey": "not-needed",
  "url": "http://ollama-server.internal:11434",
  "chatModelName": "llama3.1",
  "fieldsToExtract": "documentTitle,classification,author,date,summary,sensitivityLevel"
}

Expected output:

{
  "documentTitle": "Q1 2026 Security Assessment Report",
  "classification": "Internal Use Only",
  "author": "Security Team",
  "date": "2026-03-01",
  "summary": "Quarterly security assessment covering network infrastructure, application security, and access control audit results.",
  "sensitivityLevel": "CONFIDENTIAL"
}

Classify internal documents for archival

Use the Classify operation to automatically categorize documents for a records management workflow.

Configuration:

{
  "apiKey": "not-needed",
  "url": "http://localhost:11434",
  "chatModelName": "llama3.2",
  "categories": "FINANCIAL_REPORT,MEETING_MINUTES,POLICY_DOCUMENT,CORRESPONDENCE,TECHNICAL_SPEC,OTHER"
}

Expected output:

{
  "category": "MEETING_MINUTES",
  "confidence": 0.85
}

Configuration tips

Ensure sufficient GPU/CPU resources for acceptable response times. Larger models (70B+) require significant GPU memory.
Smaller models like llama3.2:1b or phi3 are much faster but less accurate — suitable for simple classification tasks.
Set a generous requestTimeout (60000-120000 ms) for larger models or complex prompts, as local inference can be slower than cloud APIs.
Use llama3.2 or mistral for the best balance of quality and speed on consumer hardware.
For production on-premises deployments, consider running Ollama on a dedicated GPU server accessible via the network URL parameter.
The same prompt configurations work across all AI providers, so you can develop with Ollama and deploy with a cloud provider.

Source code

bonita-connector-ai on GitHub (module bonita-connector-ai-ollama)