[BUG] problem to get all models with ollama local and remote

[kvnxp]

Problem (one or two sentences)

i have problem to get all models that have downloaded on ollama , and i use ollama on other pc too, when write ip server, cant load models from server, i see the server log that make request but on models list show nothings or nothings happends ..

show current models on local but on remote show nothings or nothing happends

model list

Context (who is affected and when)

when try to get models from ollama local or, from other pc

Reproduction steps

ollama local on macOs Thaoe
ollama remote on other macOs sequoia

Expected result

nothings happends

Actual result

show some (not all ) models on local, and show nothing on remote

Variations tried (optional)

No response

App Version

3.51.1 (7c9722b)

API Provider (optional)

Ollama

Model Used (optional)

No response

Roo Code Task Links (optional)

No response

Relevant logs or errors (optional)

[roomote-v0]

I've analyzed this issue and here's my proposed implementation plan:

Root Cause Analysis:

The Ollama model fetching code in src/api/providers/fetchers/ollama.ts has two problems that explain both symptoms (showing only some models locally, and showing nothing on remote):

1. Aggressive tools-capability filter (causes "not all models" locally): In parseOllamaModel(), models are filtered to only include those where capabilities includes "tools". Models that don't report a capabilities array at all (older Ollama versions, or models whose metadata lacks this field) have supportsTools default to false and are silently excluded. This is why only some models appear locally.

2. Silent error handling (causes "nothing" on remote): When fetching from a remote Ollama server, any network error, timeout, or unexpected response is caught silently in getOllamaModels() and returns an empty object. Additionally, webviewMessageHandler.ts only sends models to the webview if the result is non-empty (Object.keys(ollamaModels).length > 0), so the UI gets no response at all -- not even an error message.

Proposed Changes:

1. Relax the tools filter in parseOllamaModel() (src/api/providers/fetchers/ollama.ts): Instead of requiring capabilities.includes("tools"), include models even when capabilities is undefined/missing. Only exclude a model if capabilities is explicitly present AND does not include "tools". This way, models from older Ollama versions (which don't report capabilities) still show up.

2. Improve error feedback for remote connections: Update the requestOllamaModels handler in webviewMessageHandler.ts to always send a response to the webview (even if the model list is empty), and surface error information so the user gets meaningful feedback instead of silence.

3. Update tests to cover the new filtering behavior and error handling.

This approach solves "not all models showing" by being more permissive with the capability filter, and "nothing showing on remote" by ensuring errors are surfaced to the user.

Please mention @roomote in a comment saying that you approve this plan, and I'll proceed with the implementation.