Raw legal contract
AI Models & Assistants / Knowledge library
Ollama
run models locally and access supported cloud models
Why learn Ollama?
Useful for experimenting with model-backed applications through a compact runtime interface. Privacy and hardware assumptions depend on whether execution is local or hosted.
Where it came from
Ollama provides tools and APIs for running and interacting with models. Its documentation includes local execution as well as cloud-related offerings.
Build your understanding
Work through these exercises in order. Try the task before opening a reference.
- Identify where a selected model executes.
- Run a small model within available memory.
- Call the API with a fixed prompt.
- Measure response latency and memory use.
- Verify network behavior before treating a workflow as offline.
Documentation and sources
- Official documentation (docs.ollama.com)
Catalog research notes
- Research snapshot: September 9, 2026. This entry summarizes the linked primary documentation; it does not assert a latest release version.
Connect the ideas
Explore next
Real World Example
Hybrid Model Orchestration for Secure Legal Analysis
A boutique law firm handling sensitive client contracts requiring strict data residency.
The business problem
The firm needs to analyze thousands of documents for clause extraction but cannot upload sensitive legal data to public cloud providers due to strict confidentiality agreements.
How this topic solves it
The firm implements Ollama to run open-weight models locally on secure, air-gapped workstations for initial document processing, while using the Ollama API to route non-sensitive summary tasks to cloud-hosted models for advanced reasoning.
- Install the Ollama runtime on local secure workstations to host open-weight models.
- Configure the local application to interface with the Ollama API for local document parsing.
- Define a routing logic in the application layer to distinguish between sensitive data and general query tasks.
- Integrate the Ollama API client to forward non-sensitive metadata to cloud-hosted models when necessary.
- Verify that sensitive document text remains strictly within the local environment.
This flow shows how the application decides between local and cloud processing based on data sensitivity.
Logic to classify data
Secure local model execution
External reasoning service
- Document Input → routes data → Sensitivity Filter
- Sensitivity Filter → if sensitive → Local Ollama Instance
- Sensitivity Filter → if non-sensitive → Cloud Model API
Expected business outcome
The firm achieves a balance between high-security local processing for sensitive data and the advanced reasoning capabilities of cloud models for general tasks, ensuring compliance with client privacy requirements.
Limitations and checks
The primary limitation is the hardware dependency for local models; if the local workstation lacks sufficient VRAM, inference speed will degrade significantly. Developers can check this by monitoring the local process memory usage and latency during model execution.