Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/mauroperez055/infoJobs/llms.txt

Use this file to discover all available pages before exploring further.

Ollama is a local LLM runtime that InfoJobs DevBoard uses to generate job summaries entirely on your machine. Because everything runs locally, there are no API keys to manage, no per-request costs, and no data sent to a third-party service. The Express backend communicates with Ollama over localhost:11434 whenever a user requests a job summary.

​
Installation

1

Download Ollama

Go to https://ollama.com and download the installer for your operating system (macOS, Linux, or Windows). Follow the platform-specific installation instructions on the site.
2

Pull the model

Open a terminal and download the qwen2.5:3b model. This is approximately 2 GB, so the download time will depend on your connection speed.
ollama pull qwen2.5:3b
3

Start Ollama

Start the Ollama server. It will listen on localhost:11434 by default.
ollama serve
Keep this terminal open while running InfoJobs DevBoard. The backend requires the server to be reachable at that address.
4

Verify the installation

In a second terminal, confirm that the model was downloaded successfully. The qwen2.5:3b entry should appear in the output.
ollama list

​
Model options

qwen2.5:3b is the default model used by InfoJobs DevBoard, but Ollama supports several others. The table below lists models that are compatible with the same ollama.chat interface used by the backend:
ModelSizeRAM RequiredQuality
qwen2.5:3b~2GB~4GBGood (default)
llama3.2~4GB~6GBExcellent
mistral~4GB~6GBGood
phi3~2GB~4GBBasic
Choose a larger model for higher-quality summaries if your machine has the available RAM, or stick with qwen2.5:3b or phi3 on machines with limited resources.

​
Changing the model

To swap the model, open backend/routes/ai.js and update the model field in the ollama.chat call: 
const response = await ollama.chat({
  model: 'qwen2.5:3b',  // change to another model, e.g. 'llama3.2' or 'mistral'
  messages: [{ role: 'user', content: prompt }],
  stream: true,
})
After saving the file, restart the backend. The new model must already be pulled with ollama pull <model-name> before the backend can use it.

​
Troubleshooting

The backend could not reach Ollama at localhost:11434. Make sure the server is running:
ollama serve
Then verify that the model is installed:
ollama list
If qwen2.5:3b does not appear, pull it again with ollama pull qwen2.5:3b.
The summary endpoint returns a 404 with { "error": "Job Not Found" } when the requested job ID does not exist in the database. This is expected behavior — check that the job ID in the URL corresponds to a real listing.
If the browser blocks the request to the backend, verify that the Express CORS middleware is configured and applied before the AI router. Check backend/middlewares/cors.js and confirm it is imported in backend/index.js.
Larger models take more time to produce output, especially on machines without a GPU. Try switching to a lighter model:
ollama pull qwen2.5:3b
# or
ollama pull phi3
Then update the model field in backend/routes/ai.js as described in the Changing the model section above.

Build docs developers (and LLMs) love

Get started for freeTalk to us