Add support for Claude's "adaptive" and "auto" thinking modes using `output_config.effort`. Introduce support for new effort level "max" in adaptive thinking. Update thinking logic, validate model capabilities, and extend converters and handling to ensure compatibility with adaptive modes. Adjust static model data with supported levels and refine handling across translators and executors.
## Problem
When using Antigravity Claude models through CLIProxyAPI, the thinking
chain (reasoning content) does not display in the Amp client.
## Root Cause
The Amp client sends `thinking: {"type": "auto"}` in its requests,
but `ConvertClaudeRequestToAntigravity` only handled `"enabled"` and
`"adaptive"` types in its switch statement. The `"auto"` type was
silently ignored, resulting in no `thinkingConfig` being set in the
translated Gemini request. Without `thinkingConfig`, the Antigravity
API returns responses without any thinking content.
Additionally, the Antigravity API for Claude models does not support
`thinkingBudget: -1` (auto mode sentinel). It requires a concrete
positive budget value. The fix uses 128000 as the budget for "auto"
mode, which `ApplyThinking` will then normalize to stay within the
model's actual limits (e.g., capped to `maxOutputTokens - 1`).
## Changes
### internal/translator/antigravity/claude/antigravity_claude_request.go
1. **Add "auto" case** to the thinking type switch statement.
Sets `thinkingBudget: 128000` and `includeThoughts: true`.
The budget is subsequently normalized by `ApplyThinking` based
on model-specific limits.
2. **Add "auto" to hasThinking check** so that interleaved thinking
hints are injected for tool-use scenarios when Amp sends
`thinking.type="auto"`.
### internal/registry/model_definitions_static_data.go
3. **Add Thinking configuration** for `claude-sonnet-4-6`,
`claude-sonnet-4-5`, and `claude-opus-4-6` in
`GetAntigravityModelConfig()` -- these were previously missing,
causing `ApplyThinking` to skip thinking config entirely.
## Testing
- Deployed to Railway test instance (cpa-thinking-test)
- Verified via debug logging that:
- Amp sends `thinking: {"type": "auto"}`
- CPA now translates this to `thinkingConfig: {thinkingBudget: 128000, includeThoughts: true}`
- `ApplyThinking` normalizes the budget to model-specific limits
- Antigravity API receives the correct thinkingConfig
Amp-Thread-ID: https://ampcode.com/threads/T-019ca511-710d-776d-a07c-4b750f871a93
Co-authored-by: Amp <amp@ampcode.com>
- Add IAM Identity Center (IDC) authentication with CLI flags (--kiro-idc-login, --kiro-idc-start-url, --kiro-idc-region) and login flow
- Add ProfileArn auto-fetching in Execute/ExecuteStream for imported IDC accounts
- Simplify endpoint preference with map-based alias lookup and getAuthValue helper
- Redesign fingerprint as global singleton with external config and per-account deterministic generation
- Add StartURL and FingerprintConfig fields to Kiro config
- Add AgentContinuationID/AgentTaskType support in Kiro translators
- Add comprehensive tests for executor, fingerprint, SSO OIDC, and AWS helpers
- Add CLI login documentation to README
- Change RequiredFieldsByTool value type from []string to [][]string
- Outer slice = AND (all groups required); inner slice = OR (any one satisfies)
- Fix Bash entry to accept "cmd" or "command", resolving soft-truncation loop
- Update findMissingRequiredFields logic and inline docs accordingly
Move base64 image data from Claude tool_result into functionResponse.parts
as inlineData instead of outer sibling parts, preventing context bloat.
Unify all inlineData field naming to camelCase mimeType across Claude,
OpenAI, and Gemini translators. Add comprehensive edge case tests and
Gemini-side regression test for functionResponse.parts preservation.
Non-spark codex models (gpt-5.3-codex, gpt-5.2-codex) stream function call
arguments via multiple delta events followed by a done event. The done handler
unconditionally emitted the full arguments, duplicating what deltas already
streamed. This produced invalid double JSON that Claude Code couldn't parse,
causing tool calls to fail with missing parameters and infinite retry loops.
Add HasReceivedArgumentsDelta flag to track whether delta events were received.
The done handler now only emits arguments when no deltas preceded it (spark
models), while delta-based streaming continues to work for non-spark models.
Some Codex models (e.g. gpt-5.3-codex-spark) send function call arguments
in a single "done" event without preceding "delta" events. The streaming
translator only handled "delta" events, causing tool call arguments to be
lost — resulting in empty tool inputs and infinite retry loops in clients
like Claude Code.
Emit the full arguments from the "done" event as a single input_json_delta
so downstream clients receive the complete tool input.
sjson.SetRaw with an empty string produces malformed JSON (e.g. "result":}).
This happens when a Claude tool_result block has no content field, causing
functionResponseResult.Raw to be "". Guard against this by falling back to
sjson.Set with an empty string only when .Raw is empty.
The OpenAI Chat Completions translator was silently dropping
response.function_call_arguments.delta and
response.function_call_arguments.done Codex SSE events, meaning
tool call arguments were never streamed incrementally to clients.
Add proper handling mirroring the proven Claude translator pattern:
- response.output_item.added: announce tool call (id, name, empty args)
- response.function_call_arguments.delta: stream argument chunks
- response.function_call_arguments.done: emit full args if no deltas
- response.output_item.done: defensive fallback for backward compat
State tracking via HasReceivedArgumentsDelta and HasToolCallAnnounced
ensures no duplicate argument emission and correct behavior for models
like codex-spark that skip delta events entirely.
Change GenerateSearchIndicatorEvents to return [][]byte instead of []sseEvent
for consistency with BuildFallbackTextEvents and other event building functions.
Benefits:
- Consistent API across all event generation functions
- Eliminates intermediate sseEvent type conversion in caller
- Simplifies usage by returning ready-to-send SSE byte slices
This addresses the code quality feedback from PR #226 review.
Consolidate web search handler, SSE event generation, stream analysis,
and MCP HTTP I/O into the executor layer. Merge the separate
kiro_websearch_handler.go back into kiro_executor.go to align with
the single-file-per-executor convention. Translator retains only pure
data types, detection, and payload transformation.
Key changes:
- Move SSE construction (search indicators, fallback text, message_start)
from translator to executor, consistent with streamToChannel pattern
- Move MCP handler (callMcpAPI, setMcpHeaders, fetchToolDescription)
from translator to executor alongside other HTTP I/O
- Reuse applyDynamicFingerprint for MCP UA headers (eliminate duplication)
- Centralize MCP endpoint URL via BuildMcpEndpoint in translator
- Add atomic Set/GetWebSearchDescription for cross-layer tool desc cache
- Thread context.Context through MCP HTTP calls for cancellation support
- Thread usage reporter through all web search API call paths
- Add token expiry pre-check before MCP/GAR calls
- Clean up dead code (GenerateMessageID, webSearchAuthContext fp logic,
ContainsWebSearchTool, StripWebSearchTool)
Kiro/AmazonQ API requires the conversation history to start with a user message.
Some clients (e.g., OpenClaw) send conversations starting with an assistant message,
which is valid for the native Claude API but causes 'Improperly formed request' (400)
on the Kiro endpoint.
This fix detects when the first message has role=assistant and prepends a minimal
placeholder user message ('.') to satisfy the Kiro API's message ordering requirement.
Upstream error: {"message":"Improperly formed request.","reason":null}
Verified: original request returns 400, fixed request returns 200.
