fix: Auto-discover Ollama models without requiring explicit API key by spiceoogway · Pull Request #4782 · openclaw/openclaw

spiceoogway · 2026-01-30T17:24:10Z

Summary

When Ollama is running locally with models installed, they were showing as 'missing' in openclaw models list because the provider was only registered if an explicit API key was configured.

Root Cause

In src/agents/models-config.providers.ts, the resolveImplicitProviders() function only added the Ollama provider when an API key was explicitly configured via environment variables or auth profiles.

Since Ollama is a local service that doesn't require traditional API authentication (it runs on localhost:11434), this prevented auto-discovery of locally installed models.

Changes

Modified the Ollama provider registration logic to auto-discover models by attempting to query the local instance
If models are found (models.length > 0), the provider is registered automatically
Uses a placeholder API key 'ollama-local' for local instances
Still respects explicit configuration if provided (backward compatible)

Testing

✅ Linted the changed file: no warnings or errors
✅ Aligns with patterns used for other local/unauthenticated providers (e.g., qwen-portal with OAuth placeholder)

Expected Behavior After Fix

User runs ollama pull deepseek-r1:latest
User runs openclaw models list
Ollama models are now listed as available (not missing)
User can set default model: openclaw models set ollama/deepseek-r1:latest
Chat can successfully use the Ollama model

Impact

Low risk: Only affects Ollama provider discovery
High value: Enables out-of-the-box Ollama support without manual configuration
Backward compatible: Existing configurations with explicit API keys still work

Note: This change makes OpenClaw's Ollama integration work the same way as other local model servers - auto-discover and register if running.

Greptile Overview

Greptile Summary

This PR changes implicit provider resolution to auto-register Ollama when a local Ollama instance responds with at least one installed model. It does so by always building the Ollama provider (which triggers /api/tags discovery against http://127.0.0.1:11434) and, if any models are returned, registering the provider with either an explicit env/auth-profile key or a placeholder ollama-local key.

This fits into the existing resolveImplicitProviders() pattern, which opportunistically adds providers based on available credentials/profiles and discovery results. The main behavior change is that Ollama discovery is no longer strictly gated by explicit configuration; it can now be activated by local state alone.

Confidence Score: 4/5

This PR is likely safe to merge, but it changes Ollama discovery semantics and may introduce noisy localhost probing/logging in common cases.
The code change is small and localized, but it alters when discovery runs (potentially affecting users who rely on opt-in behavior or explicit config) and can emit warnings during normal “Ollama not running/no models” scenarios.
src/agents/models-config.providers.ts

Context used:

Context from dashboard - CLAUDE.md (source)
Context from dashboard - AGENTS.md (source)

Fixes openclaw#4544 When Ollama is running locally with models installed, they were showing as 'missing' in 'openclaw models list' because the provider was only registered if an explicit API key was configured. Since Ollama is a local service that doesn't require authentication, this change: - Auto-discovers Ollama models by attempting to query the local instance - Registers the provider if models are found - Uses a placeholder API key 'ollama-local' for local instances - Still respects explicit configuration if provided This aligns with how other local/unauthenticated providers work (e.g., qwen-portal with OAuth placeholder). Users can now use Ollama models immediately after installing Ollama and pulling models, without needing to configure authentication.

greptile-apps

_{1 file reviewed, 2 comments}

_{Edit Code Review Agent Settings | Greptile}

greptile-apps · 2026-02-03T02:24:55Z

src/agents/models-config.providers.ts

+  // Ollama provider - auto-discover if running locally
+  const ollamaProvider = await buildOllamaProvider();
+  if (ollamaProvider.models.length > 0) {
+    const ollamaKey =
+      resolveEnvApiKeyVarName("ollama") ??
+      resolveApiKeyFromProfiles({ provider: "ollama", store: authStore }) ??
+      "ollama-local"; // Placeholder for local Ollama instance
+    providers.ollama = { ...ollamaProvider, apiKey: ollamaKey };


[P1] Ollama discovery now runs even when explicitly configured; may break “explicit config disables auto-discovery” contract

resolveImplicitProviders() will always call buildOllamaProvider() (which hits http://127.0.0.1:11434/api/tags) before checking for any explicit OLLAMA_API_KEY/auth profile. If a user has explicitly configured Ollama for a remote host or expects discovery to be opt-in, this adds an unexpected localhost probe and can cause the implicit provider to appear/disappear based purely on local state. Consider gating discovery behind “no explicit provider entry” (or behind an explicit opt-in flag) to preserve the documented behavior.

Also appears inconsistent with docs in docs/providers/ollama.md that say implicit discovery is opt-in via OLLAMA_API_KEY and skipped when models.providers.ollama is explicitly set.

Prompt To Fix With AI

This is a comment left during a code review. Path: src/agents/models-config.providers.ts Line: 456:463 Comment: [P1] **Ollama discovery now runs even when explicitly configured; may break “explicit config disables auto-discovery” contract** `resolveImplicitProviders()` will always call `buildOllamaProvider()` (which hits `http://127.0.0.1:11434/api/tags`) before checking for any explicit `OLLAMA_API_KEY`/auth profile. If a user has explicitly configured Ollama for a remote host or expects discovery to be opt-in, this adds an unexpected localhost probe and can cause the implicit provider to appear/disappear based purely on local state. Consider gating discovery behind “no explicit provider entry” (or behind an explicit opt-in flag) to preserve the documented behavior. Also appears inconsistent with docs in `docs/providers/ollama.md` that say implicit discovery is opt-in via `OLLAMA_API_KEY` and skipped when `models.providers.ollama` is explicitly set. How can I resolve this? If you propose a fix, please make it concise.

greptile-apps · 2026-02-03T02:25:00Z

Additional Comments (1)

src/agents/models-config.providers.ts
[P2] Noisy warnings during normal “not running / no models” case

discoverOllamaModels() logs warnings for expected situations (!response.ok, empty models). Since resolveImplicitProviders() now probes Ollama by default, these warnings can show up for users who don’t use Ollama (or haven’t pulled models yet) and make openclaw models list look unhealthy. Consider downgrading these to debug-level logging or only warning when the user explicitly enabled Ollama discovery.

_{Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!}

Prompt To Fix With AI

This is a comment left during a code review.
Path: src/agents/models-config.providers.ts
Line: 100:110

Comment:
[P2] **Noisy warnings during normal “not running / no models” case**

`discoverOllamaModels()` logs warnings for expected situations (`!response.ok`, empty `models`). Since `resolveImplicitProviders()` now probes Ollama by default, these warnings can show up for users who don’t use Ollama (or haven’t pulled models yet) and make `openclaw models list` look unhealthy. Consider downgrading these to debug-level logging or only warning when the user explicitly enabled Ollama discovery.

<sub>Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!</sub>

How can I resolve this? If you propose a fix, please make it concise.

…4544) Ollama is a local service that doesn't require authentication, but the provider was only registered when OLLAMA_API_KEY was set. This caused locally installed models to show as "missing" in openclaw models list. Changes: - Always attempt Ollama model discovery via /api/tags - Register provider with placeholder key "ollama-local" when models found - Add quiet flag to suppress warnings during implicit auto-discovery, so users without Ollama won't see noisy console output - Preserve existing behavior when API key is explicitly configured Fixes openclaw#4544 Ref openclaw#4782

steipete · 2026-02-24T04:44:21Z

AI-assisted stale-fix triage closure.

Why closing:

Superseded by fix: Auto-discover Ollama models without requiring API key #20872 for [Bug]: Cannot change model to Ollama Deepseek-r1:latest #4544.
fix: Auto-discover Ollama models without requiring API key #20872 keeps the same auto-discovery fix and adds quieter/noise-safe behavior plus stronger regression coverage.

Result:

Closed as stale superseded path; keep fix: Auto-discover Ollama models without requiring API key #20872 as canonical PR for this issue.

If still relevant:

Reopen with a distinct uncovered Ollama discovery case not addressed by fix: Auto-discover Ollama models without requiring API key #20872.

steipete · 2026-02-24T04:44:21Z

Closed after AI-assisted stale-fix triage; see prior comment for details.

…4544) Ollama is a local service that doesn't require authentication, but the provider was only registered when OLLAMA_API_KEY was set. This caused locally installed models to show as "missing" in openclaw models list. Changes: - Always attempt Ollama model discovery via /api/tags - Register provider with placeholder key "ollama-local" when models found - Add quiet flag to suppress warnings during implicit auto-discovery, so users without Ollama won't see noisy console output - Preserve existing behavior when API key is explicitly configured Fixes openclaw#4544 Ref openclaw#4782

openclaw-barnacle bot added channel: nostr Channel integration: nostr agents Agent runtime and tooling labels Jan 30, 2026

spiceoogway force-pushed the fix/issue-4544-ollama-auto-discovery branch 2 times, most recently from a893a43 to 0ea97fc Compare January 30, 2026 17:25

openclaw-barnacle bot removed the channel: nostr Channel integration: nostr label Jan 30, 2026

spiceoogway force-pushed the fix/issue-4544-ollama-auto-discovery branch from 0ea97fc to 8ecc1c6 Compare January 30, 2026 17:26

spiceoogway mentioned this pull request Jan 31, 2026

fix: Auto-discover Ollama models without requiring API key #4667

Closed

spiceoogway force-pushed the fix/issue-4544-ollama-auto-discovery branch from 8ecc1c6 to b888c82 Compare January 31, 2026 21:25

Copilot AI mentioned this pull request Feb 2, 2026

Integrate Ollama auto-discovery and installer charlieduzstuf/openclaw#1

Merged

greptile-apps bot reviewed Feb 3, 2026

View reviewed changes

thewilloftheshadow force-pushed the main branch from bfc1ccb to f92900f Compare February 15, 2026 18:46

marcodelpin mentioned this pull request Feb 19, 2026

fix: Auto-discover Ollama models without requiring API key #20872

Closed

4 tasks

steipete closed this Feb 24, 2026

vincentkoc mentioned this pull request Feb 27, 2026

fix: harden Ollama autodiscovery and warning behavior #29201

Merged

18 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

fix: Auto-discover Ollama models without requiring explicit API key#4782

fix: Auto-discover Ollama models without requiring explicit API key#4782
spiceoogway wants to merge 1 commit intoopenclaw:mainfrom
spiceoogway:fix/issue-4544-ollama-auto-discovery

spiceoogway commented Jan 30, 2026 •

edited by greptile-apps bot

Loading

Uh oh!

greptile-apps bot left a comment

Uh oh!

greptile-apps bot Feb 3, 2026

Uh oh!

greptile-apps bot commented Feb 3, 2026

Uh oh!

steipete commented Feb 24, 2026

Uh oh!

steipete commented Feb 24, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

spiceoogway commented Jan 30, 2026 • edited by greptile-apps bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Root Cause

Changes

Testing

Expected Behavior After Fix

Impact

Greptile Overview

Greptile Summary

Confidence Score: 4/5

Uh oh!

greptile-apps bot left a comment

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot Feb 3, 2026

Choose a reason for hiding this comment

Uh oh!

greptile-apps bot commented Feb 3, 2026

Uh oh!

steipete commented Feb 24, 2026

Uh oh!

steipete commented Feb 24, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

spiceoogway commented Jan 30, 2026 •

edited by greptile-apps bot

Loading