fix: expand AUTODETECT for tabAutocompleteModel in JSON config path

[riyonp23]

Fixes #12400

Description

The JSON config path (config.json) never expanded an AUTODETECT placeholder in tabAutocompleteModel. When users configured model: "AUTODETECT" for autocomplete, the placeholder LLM went straight into modelsByRole.autocomplete unchanged, causing Continue to send the literal string "AUTODETECT" to the provider at request time. The YAML config path was already handling this correctly.

Fixed by mirroring the existing AUTODETECT expansion logic from the config.models loop into the tabAutocompleteModels block in core/config/load.ts. Also exported intermediateToFinalConfig for testability.

AI Code Review

• Team members only: AI review runs automatically when PR is opened or marked ready for review
• Team members can also trigger a review by commenting @continue-review

Checklist

• I've read the contributing guide
• The relevant docs, if any, have been updated or created
• The relevant tests, if any, have been updated or created

Screen recording or screenshot

Not applicable — this is a backend config parsing fix with no UI changes.

Tests

Added core/config/load.vitest.ts with 3 new cases:

1. AUTODETECT expands correctly — asserts no AUTODETECT entry remains in modelsByRole.autocomplete and both expanded models are present with isFromAutoDetect: true
2. Non-AUTODETECT config passes through unchanged
3. listModels() rejection — asserts the AUTODETECT placeholder does not leak through and console.warn was emitted

All 3 pass. Existing core/config/yaml/models.vitest.ts (8 cases) still passes — YAML path unaffected.

---

Note: While working on this I noticed a pre-existing unrelated bug in load.ts:316–334 — const models = modelNames.map(...) shadows the outer models: BaseLLM[] in the CustomLLM AUTODETECT branch, so expanded models never reach the outer array. Left it untouched to keep this PR focused. Worth a separate issue.

---

Summary by cubic

Fixes the JSON config path so tabAutocompleteModel expands AUTODETECT instead of sending the literal string to providers. Mirrors YAML behavior and adds coverage to prevent regressions.

• Bug Fixes

  • Expand AUTODETECT for tabAutocompleteModel in the JSON path (built-in and CustomLLM), correctly populating modelsByRole.autocomplete.
  • On listModels() failure, drop the placeholder and warn, ensuring "AUTODETECT" never reaches a provider.

• Refactors

  • Exported intermediateToFinalConfig for testability.

^{Written for commit effcdce. Summary will update on new commits. Review in cubic}

[github-actions]

All contributors have signed the CLA ✍️ ✅
_{Posted by the CLA Assistant Lite bot.}

[riyonp23]

I have read the CLA Document and I hereby sign the CLA

[cubic-dev-ai]

1 issue found across 2 files

Prompt for AI agents (unresolved issues)

Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.


<file name="core/config/load.ts">

<violation number="1" location="core/config/load.ts:379">
P2: AUTODETECT expansion logic is duplicated from the earlier `config.models` loop, and drift already exists between the two copies (e.g., the pre-existing CustomLLMClass shadowing bug in `config.models` is avoided here, but other behavioral differences like free-trial inline filtering remain). Extract a shared helper to prevent future regressions.</violation>
</file>

<sub>Reply with feedback, questions, or to request a fix.

Re-trigger cubic</sub>

[cubic-dev-ai]

P2: AUTODETECT expansion logic is duplicated from the earlier config.models loop, and drift already exists between the two copies (e.g., the pre-existing CustomLLMClass shadowing bug in config.models is avoided here, but other behavioral differences like free-trial inline filtering remain). Extract a shared helper to prevent future regressions.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At core/config/load.ts, line 379:

<comment>AUTODETECT expansion logic is duplicated from the earlier `config.models` loop, and drift already exists between the two copies (e.g., the pre-existing CustomLLMClass shadowing bug in `config.models` is avoided here, but other behavioral differences like free-trial inline filtering remain). Extract a shared helper to prevent future regressions.</comment>

<file context>
@@ -372,15 +372,75 @@ async function intermediateToFinalConfig({
+            return;
+          }
+
+          if (llm.model === "AUTODETECT") {
+            try {
+              const modelNames = await llm.listModels();
</file context>