Title

Fix #12770 - Set valid Antigravity DEFAULT_MODEL and trim leading whitespace in BlueBubbles replies

Body

Summary

• Problem: Antigravity was using an invalid DEFAULT_MODEL, and BlueBubbles replies could contain unintended leading whitespace when blocks were coalesced or dispatched.

• Why it matters: The invalid model ID caused runtime failures during model resolution. Leading whitespace in outbound replies could break validation logic and produce inconsistent message formatting.

• What changed:

  • Updated DEFAULT_MODEL to google-antigravity/claude-opus-4-5-thinking.
  • Trimmed leading whitespace when coalescing reply blocks.
  • Trimmed leading whitespace before validation and dispatch in BlueBubbles send path.
  • Updated English testing documentation to reference the new Antigravity model.

• What did NOT change (scope boundary):

  • No changes to model selection logic beyond the default value.
  • No behavioral changes to reply formatting other than leading whitespace normalization.
  • No API contract or integration flow changes.

Change Type (select all)

• Bug fix
• Feature
• Refactor
• Docs
• Security hardening
• Chore/infra

Scope (select all touched areas)

• Gateway / orchestration
• Skills / tool execution
• Auth / tokens
• Memory / storage
• Integrations
• API / contracts
• UI / DX
• CI/CD / infra

Linked Issue/PR

• Closes BlueBubbles: Leading newlines not trimmed from message text #12770
• Related #

User-visible / Behavior Changes

• Antigravity now defaults to a valid model ID.
• BlueBubbles outbound messages no longer include unintended leading whitespace.

Security Impact (required)

• New permissions/capabilities? No
• Secrets/tokens handling changed? No
• New/changed network calls? No
• Command/tool execution surface changed? No
• Data access scope changed? No

No security surface changes introduced.

Repro + Verification

Environment

• OS: Windows 11 / WSL2 Ubuntu 24.04
• Runtime/container: Node (project default runtime)
• Model/provider: Antigravity (Google provider)
• Integration/channel: BlueBubbles
• Relevant config (redacted): Default Antigravity configuration

Steps (Before Fix)

1. Start with default Antigravity configuration.
2. Trigger model resolution without explicit override.
3. Observe model resolution failure due to invalid DEFAULT_MODEL.

For BlueBubbles:

1. Send a multi-block reply containing leading newline(s).
2. Observe outbound message including unintended leading whitespace.

Expected

• Antigravity resolves a valid default model without runtime errors.
• BlueBubbles replies are normalized (no leading whitespace).

Actual (Before Fix)

• Runtime failure when resolving default Antigravity model.
• Outbound BlueBubbles messages could contain leading whitespace.

Evidence

• Trace/log snippets
• Failing test/log before + passing after
• Screenshot/recording
• Perf numbers (if relevant)

Verified via runtime logs and manual integration testing.

Human Verification (required)

• Verified scenarios:

  • Default Antigravity model loads successfully.
  • Multi-block replies are coalesced without leading whitespace.
  • BlueBubbles send path trims text before validation and dispatch.

• Edge cases checked:

  • Multiple leading newline characters.
  • Empty string after trimming.
  • Normal single-block replies unaffected.

• What you did NOT verify:

  • Full CI matrix across all providers.
  • Non-English doc variants (only English testing docs updated).

Compatibility / Migration

• Backward compatible? Yes
• Config/env changes? No
• Migration needed? No

No migration steps required.

Failure Recovery (if this breaks)

• How to disable/revert this change quickly:
  Revert this PR or restore previous DEFAULT_MODEL value.

• Files/config to restore:

  • extensions/google-antigravity-auth/index.ts
  • BlueBubbles send path files

• Known bad symptoms reviewers should watch for:

  • Model resolution failures.
  • Unexpected message formatting differences beyond leading whitespace.

Risks and Mitigations

• Risk: Trimming could remove intentional formatting in rare cases.

  • Mitigation: Only leading whitespace is trimmed; internal formatting remains untouched.

Greptile Summary

Updates the Antigravity DEFAULT_MODEL from claude-opus-4-6-thinking to claude-opus-4-5-thinking and adds trimStart() normalization to BlueBubbles outbound text to strip leading whitespace before validation and dispatch. The model change aligns with the existing forward-compat fallback in src/agents/model-forward-compat.ts.

• extensions/google-antigravity-auth/index.ts: DEFAULT_MODEL constant updated to valid model ID.
• extensions/bluebubbles/src/send.ts: trimStart() added inside sendMessageBlueBubbles, normalizing leading whitespace for all callers (channel, actions, monitor-processing, media-send).
• extensions/bluebubbles/src/channel.ts: trimStart() also added in the sendText outbound path before calling sendMessageBlueBubbles — this is redundant with the send.ts change but not harmful.
• docs/help/testing.md: Model list updated on line 264, but two command-line examples on lines 241 and 269 still reference the old claude-opus-4-6-thinking model name. These should be updated for consistency.

Confidence Score: 3/5

• Low-risk changes overall, but the docs have incomplete model name updates that will cause confusion when running smoke tests.
• The code changes in send.ts, channel.ts, and index.ts are straightforward and correct. However, the docs update in testing.md is incomplete — two command-line examples still reference the old invalid model name, which undermines the purpose of this PR. Additionally, the zh-CN translation docs also still reference the old model name (though these are noted as generated and out of scope).
• Pay close attention to docs/help/testing.md — lines 241 and 269 still reference the old model name claude-opus-4-6-thinking and need to be updated to match line 264.

_{Last reviewed commit: 36d193e}

_{(2/5) Greptile learns from your feedback when you react with thumbs up/down!}