Resolve conflicts:
- streamingHandler.js: adopt upstreamResponseHeaders while keeping 0-token detail row avoidance
- capabilities.js: preserve user-asserted caps and globalThis slots without local caching of catalogSource
- AddCustomModelModal.js & providers/[id]/page.js: wire STT transport marker with custom model edits/assertions
- models/custom/route.js & aliasRepo.js: persist custom model transport and invalidate user caps
- usageRepo.js: key byApiKey live stats by full API key and keep tail in maskApiKey
- UsageStats.js: lazy load charts dynamically
Anthropic's API-level refusal (streaming classifier / ToS) ends the stream
with stop_reason "refusal", stop_details carrying the reason, zero output
tokens and no content blocks. Map refusal to content_filter in both
directions, surface stop_details.explanation as message text, and add
CLAUDE_STOP.REFUSAL to schema.
Map upstream Chat Completions usage to the Responses API shape and attach it to response.completed. Capture chunk.usage before the empty-choices guard so the usage-only trailer chunk survives, and defer completion to flushEvents() when usage is not yet known — only on the direct openai:openai-responses route, since a pivoted stream never reaches flushEvents. Fixes#3432.
Do not collapse consecutive underscores in uniqueName so mcp__server__tool is sent intact to Kiro, attach reverse map on request translation, and restore client tool names in responses.
Forward images inside tool_result to OpenAI and Kiro upstreams via following user messages, restore original client tool names on Kiro responses via _toolNameMap, and preserve thinking display settings across translations.
The merged Qoder work also rewrote shared translator/handler code so that
/v1/responses clients got token usage on response.completed. That changed
behaviour for every provider, not just Qoder: proxies saw input tokens
rise by the 2000-token context buffer, and the plain token mapping was
replaced by one that always adds input_tokens_details.
A probe confirms the Qoder benefit does not depend on those edits: the
executor's coalescer already emits one include_usage-style finish chunk, so
a Claude client receives input_tokens and cache_read_input_tokens with
every shared file at its original state. Only the Responses path relies on
the shared translator, and that path has no Qoder-owned seam to put it in.
Reverts the shared files to their pre-PR state and drops the Responses
usage test. The Cline envelope unwrap in nonStreamingHandler.js, which
landed after the PR in the same file, is kept.
- Coalesce Qoder's empty finish-in-delta frame with the later choices:[] usage
frame so OpenAI and Claude clients receive prompt_tokens, completion_tokens
and cache-hit tokens (the dashboard already saw them)
- Upload inlined images through /api/v2/image/upload like qodercli, and stub
oversized non-image files instead of stuffing 30MB+ data URIs into
agent_chat_generation
- Emit response.completed -> response.usage for chat-native upstreams so
/v1/responses clients (Codex CLI, sub2api) no longer log 0/0/0
- Keep Claude message_delta.usage working when usage arrives without choices[0]
- Escalate to the smallest advertised Qoder context tier (200K/400K/1M) when
the estimated prompt no longer fits max_input_tokens
- Pass apiKey for PAT connections and list hidden enable:false catalog keys
from /v1/models
- Add muse-spark-1.3-contributor as responses-only model on OpenCode Go with dedicated executor
- Key Responses→chat streaming tool calls by item_id to prevent parallel tool calls merging into index 0
- Standardize tool coercions and call_id clamping in Responses API translation
Some providers (e.g. codebuddy/cbcn) attach an empty tool_calls array to every streaming chunk. An empty array is truthy in JS, so the guard 'if (delta.tool_calls)' closed the message on the first content token and emitted response.output_text.done early, dropping the remaining deltas. Guard on a non-empty array; finish_reason still closes the message and real tool calls still close it before emitting function_call items.
fixes#3234
Codex Responses Lite clients routed to a chat-native OpenAI-compatible
provider lost tool use in three places: non-streaming Chat responses
leaked the raw chat.completion envelope instead of Responses output
items, internal reasoning continuity fields leaked into the outbound
Chat body causing some upstreams to reject the request, and the
Responses to Chat request translator ignored additional_tools,
custom_tool_call, and custom_tool_call_output items entirely.
Also fixes apiType (chat vs responses) for openai-compatible nodes
being resolved from the immutable provider ID instead of the stored
node config, so editing a node's API Type had no runtime effect.
Upstream emits AI SDK v5 {"type":"error"} events inside an HTTP 200 stream.
The translator turned them into fake success content ([CommandCode error: ...]
+ finish_reason stop), so account/model fallback never fired and logs showed
Status: success.
- translator: error events now emit an OpenAI-shaped error chunk (chunk.error)
instead of content; parseSSEToOpenAIResponse already detects chunk?.error
- executor: peek the first events before committing the response; an early
error event returns 502 so fallback runs before any byte reaches the client
Normalize every provider to one cache-inclusive convention via
canonicalizeUsage() before persist, and price cached + cache_creation as
subsets of prompt_tokens in calculateCostFromTokens() to stop
double-counting. usageRepo now delegates cost math to a single source.
Surface Cached tokens/cost across dashboard (overview, tokens, cost,
details). Merge Claude message_start cache with message_delta output so
cache counts survive. Compatible LLM nodes now allow multiple API-key
connections (key pool).
Co-authored-by: Cursor <cursoragent@cursor.com>
Gemini response translation wrote OpenAI-shaped bookkeeping into the
shared state.toolCalls map, which the downstream openai-to-claude
translator uses for Claude block metadata. That pre-population skipped
blockIndex creation, so Anthropic input_json_delta events lost index.
Track Gemini function calls via state.geminiToolCallCount instead,
leaving state.toolCalls clean for the Claude translator.
Closes#2218
Co-authored-by: Cursor <cursoragent@cursor.com>
Adds long-lived API-key (ksk_) authentication for Kiro/AWS CodeWhisperer
and a direct claude:kiro / kiro:claude translation route that avoids the
lossy OpenAI two-hop pivot.
- translator: claude-to-kiro request + kiro-to-claude response translators,
registered on the exact source:target pair (direct route ahead of the
OpenAI pivot in index.js). claude-to-kiro uses shared schema constants
(ROLE/CLAUDE_BLOCK/DEFAULT_IMAGE_MIME) per app convention.
- auth: POST /api/oauth/kiro/api-key imports + validates a key via
ListAvailableProfiles, persists authMethod="api_key" (no refresh token).
- executor: send tokentype: API_KEY header and try *.amazonaws.com hosts
first for api-key creds; OAuth keeps kiro.dev first.
- fix: never inject the default placeholder profileArn for api-key auth
(CodeWhisperer 403s an ARN not owned by the key's account).
- ui: API Key method in the Kiro connect modal; surface api-key accounts
on the Quota Tracker and provider count.
- stream: env-overridable TTFT vs stall timeouts + Kiro keepalive frame.
- tests: claude-kiro-direct + kiro-profile-arn (11 tests).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
Add helpers/usageHelper.js for conditional prompt/completion token details.
Apply to gemini/codex/claude response translators; keep each provider's token
math intact. No behavior change; golden + gate: no regression.
Co-authored-by: Cursor <cursoragent@cursor.com>
Add helpers/chunkBuilder.js; apply to claude/gemini/kiro/ollama/commandcode/
openai-responses response translators. Caller supplies id/created/model so each
keeps exact id-generation + usage semantics. Extend golden response stream to
openai-responses (codex). No behavior change; gate: no regression.
Co-authored-by: Cursor <cursoragent@cursor.com>
* fix: sanitize Read tool args to prevent retry loops from non-Anthropic models
* fix: sanitize invalid Read pages from tool args
Non-Anthropic models sometimes emit optional Read args like pages: "" for
non-PDF files, which Claude Code rejects before the tool runs. Drop invalid
pages values, keep valid PDF page ranges, and coerce numeric string bounds
before clamping limit/offset.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Some OpenAI-compatible providers emit optional string tool parameters
as empty strings (e.g. pages: "") instead of omitting them. Claude
Code rejects pages: "" as invalid, breaking the Read tool for
non-PDF files routed through 9router.
Add sanitizeToolArguments() that parses tool-call arguments and
removes known optional empty-string fields before emitting
input_json_delta back to Claude format. Currently handles the
Read tool pages field specifically.
Includes regression test.
Fixes#1278
Co-authored-by: JoJo <noreply@github.com>
Replace empty reasoning_content with explicit </think> closing tag when exiting thinking block to properly signal end of reasoning section in streaming responses.
- Encode thoughtSignature into tool_call.id using _TSIG_ delimiter and base64url
- Decode _TSIG_ on request to restore thoughtSignature for Gemini multi-turn thinking
- Track pendingThoughtSignature across parts for deferred signature attachment
- Add LocalMutex (2-layer locking) to prevent ELOCKED on concurrent DB access
- Increase lockfile retries from 5 to 15 for multi-process robustness
- Restore db.json seed on first run to prevent ENOENT on lockfile.lock
- Use process.env.BASE_URL fallback in models test route
- Remove gemini-3-flash-lite-preview from provider models
Co-authored-by: kwanLeeFrmVi <quanle96@outlook.com>
Closes#450
Made-with: Cursor
