@fudrouter/fsrouter 0.6.62 → 0.6.63
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/bin/copyJs.js +0 -4
- package/dist/lib/oauth/constants/oauth.js +2 -2
- package/dist/lib/oauth/providers.js +1 -1
- package/dist/routes/providers/[id]/test/testUtils.js +3 -3
- package/dist/shared/constants/models.js +2 -2
- package/dist/sse/handlers/chat.js +10 -10
- package/dist/sse/handlers/embeddings.js +3 -3
- package/dist/sse/handlers/fetch.js +4 -4
- package/dist/sse/handlers/imageGeneration.js +5 -5
- package/dist/sse/handlers/search.js +4 -4
- package/dist/sse/handlers/stt.js +3 -3
- package/dist/sse/handlers/tts.js +4 -4
- package/dist/sse/handlers/videoGeneration.js +4 -4
- package/dist/sse/services/auth.js +2 -2
- package/dist/sse/services/model.js +1 -1
- package/dist/sse/services/tokenRefresh.js +3 -3
- package/open-sse/AGENTS.md +39 -0
- package/open-sse/config/anthropicHeaders.js +87 -0
- package/open-sse/config/appConstants.js +238 -0
- package/open-sse/config/audioRegistry.js +506 -0
- package/open-sse/config/cliFingerprints.js +304 -0
- package/open-sse/config/codexInstructions.js +119 -0
- package/open-sse/config/codexQuotaScopes.js +42 -0
- package/open-sse/config/constants.js +4 -0
- package/open-sse/config/contextEditing.js +68 -0
- package/open-sse/config/credentialLoader.js +83 -0
- package/open-sse/config/defaultThinkingSignature.js +12 -0
- package/open-sse/config/embeddingRegistry.js +264 -0
- package/open-sse/config/errorConfig.js +96 -0
- package/open-sse/config/freeModelCatalog.data.js +121 -0
- package/open-sse/config/freeModelCatalog.js +90 -0
- package/open-sse/config/freeTierCatalog.js +81 -0
- package/open-sse/config/googleTtsLanguages.js +62 -0
- package/open-sse/config/imageParamDefs.js +360 -0
- package/open-sse/config/imageRegistry.js +493 -0
- package/open-sse/config/kiroConstants.js +268 -0
- package/open-sse/config/mediaConfig.js +27 -0
- package/open-sse/config/mediaServiceKinds.js +71 -0
- package/open-sse/config/models.js +13 -0
- package/open-sse/config/moderationRegistry.js +71 -0
- package/open-sse/config/musicRegistry.js +99 -0
- package/open-sse/config/ocrRegistry.js +61 -0
- package/open-sse/config/ollamaModels.js +19 -0
- package/open-sse/config/providerErrorRules.js +114 -0
- package/open-sse/config/providerFieldStrips.js +35 -0
- package/open-sse/config/providerHeaderProfiles.js +191 -0
- package/open-sse/config/providerModels.js +106 -0
- package/open-sse/config/providerRegistry.js +16 -0
- package/open-sse/config/providers.js +19 -0
- package/open-sse/config/registryUtils.js +106 -0
- package/open-sse/config/rerankRegistry.js +177 -0
- package/open-sse/config/runtimeConfig.js +96 -0
- package/open-sse/config/runway.js +39 -0
- package/open-sse/config/searchRegistry.js +153 -0
- package/open-sse/config/toolCloaking.js +199 -0
- package/open-sse/config/ttsModels.js +129 -0
- package/open-sse/config/videoRegistry.js +207 -0
- package/open-sse/executors/adapta-web.js +435 -0
- package/open-sse/executors/antigravity.js +534 -0
- package/open-sse/executors/antigravityUpstreamError.js +13 -0
- package/open-sse/executors/auggie.js +437 -0
- package/open-sse/executors/azure-openai.js +32 -0
- package/open-sse/executors/azure.js +57 -0
- package/open-sse/executors/base.js +181 -0
- package/open-sse/executors/bedrock.js +633 -0
- package/open-sse/executors/blackbox-web.js +578 -0
- package/open-sse/executors/chatgpt-web.js +2042 -0
- package/open-sse/executors/chipotle.js +357 -0
- package/open-sse/executors/claude-web-with-auto-refresh.js +73 -0
- package/open-sse/executors/claude-web.js +586 -0
- package/open-sse/executors/cliproxyapi.js +296 -0
- package/open-sse/executors/cloudflare-ai.js +46 -0
- package/open-sse/executors/codebuddy-cn.js +40 -0
- package/open-sse/executors/codex.js +1324 -0
- package/open-sse/executors/commandCode.js +558 -0
- package/open-sse/executors/commandcode.js +88 -0
- package/open-sse/executors/copilot-m365-connection.js +176 -0
- package/open-sse/executors/copilot-m365-frames.js +192 -0
- package/open-sse/executors/copilot-m365-web.js +299 -0
- package/open-sse/executors/copilot-web.js +580 -0
- package/open-sse/executors/cursor.js +795 -0
- package/open-sse/executors/deepseek-web-with-auto-refresh.js +150 -0
- package/open-sse/executors/deepseek-web.js +875 -0
- package/open-sse/executors/default.js +430 -0
- package/open-sse/executors/devin-cli.js +351 -0
- package/open-sse/executors/doubao-web.js +550 -0
- package/open-sse/executors/duckduckgo-web.js +629 -0
- package/open-sse/executors/forceResponsesUpstream.js +33 -0
- package/open-sse/executors/gemini-business.js +320 -0
- package/open-sse/executors/gemini-cli.js +89 -0
- package/open-sse/executors/gemini-web.js +297 -0
- package/open-sse/executors/github.js +379 -0
- package/open-sse/executors/gitlab.js +569 -0
- package/open-sse/executors/glm.js +361 -0
- package/open-sse/executors/grok-cli.js +119 -0
- package/open-sse/executors/grok-web.js +345 -0
- package/open-sse/executors/huggingchat.js +467 -0
- package/open-sse/executors/iflow.js +108 -0
- package/open-sse/executors/index.js +78 -0
- package/open-sse/executors/inner-ai.js +527 -0
- package/open-sse/executors/kimchi.js +123 -0
- package/open-sse/executors/kimi-web.js +325 -0
- package/open-sse/executors/kimi.js +97 -0
- package/open-sse/executors/kiro.js +503 -0
- package/open-sse/executors/kiroThinking.js +44 -0
- package/open-sse/executors/lmarena.js +395 -0
- package/open-sse/executors/mimo-free.js +167 -0
- package/open-sse/executors/mimocode.js +392 -0
- package/open-sse/executors/ninerouter.js +155 -0
- package/open-sse/executors/nlpcloud.js +438 -0
- package/open-sse/executors/ollama-local.js +14 -0
- package/open-sse/executors/opencode-go.js +41 -0
- package/open-sse/executors/opencode.js +32 -0
- package/open-sse/executors/perplexity-web.js +581 -0
- package/open-sse/executors/poe-web.js +133 -0
- package/open-sse/executors/pollinations.js +95 -0
- package/open-sse/executors/puter.js +31 -0
- package/open-sse/executors/qoder.js +448 -0
- package/open-sse/executors/qwen-web.js +372 -0
- package/open-sse/executors/qwen.js +129 -0
- package/open-sse/executors/t3-chat-web.js +417 -0
- package/open-sse/executors/theoldllm.js +290 -0
- package/open-sse/executors/trae.js +418 -0
- package/open-sse/executors/v0-vercel-web.js +141 -0
- package/open-sse/executors/venice-web.js +142 -0
- package/open-sse/executors/vertex.js +131 -0
- package/open-sse/executors/vertexMedia.js +236 -0
- package/open-sse/executors/windsurf.js +560 -0
- package/open-sse/executors/xai.js +46 -0
- package/open-sse/executors/xiaomi-tokenplan.js +19 -0
- package/open-sse/executors/yuanbao-web.js +385 -0
- package/open-sse/executors/zed-hosted.js +281 -0
- package/open-sse/executors/zenmux-free.js +247 -0
- package/open-sse/handlers/audioSpeech.js +759 -0
- package/open-sse/handlers/audioTranscription.js +395 -0
- package/open-sse/handlers/audioTranslation.js +181 -0
- package/open-sse/handlers/chatCore/nonStreamingHandler.js +300 -0
- package/open-sse/handlers/chatCore/requestDetail.js +102 -0
- package/open-sse/handlers/chatCore/sseToJsonHandler.js +231 -0
- package/open-sse/handlers/chatCore/streamingHandler.js +95 -0
- package/open-sse/handlers/chatCore.js +404 -0
- package/open-sse/handlers/embeddingProviders/_base.js +4 -0
- package/open-sse/handlers/embeddingProviders/gemini.js +54 -0
- package/open-sse/handlers/embeddingProviders/index.js +23 -0
- package/open-sse/handlers/embeddingProviders/openai.js +39 -0
- package/open-sse/handlers/embeddingProviders/openaiCompatNode.js +13 -0
- package/open-sse/handlers/embeddingsCore.js +126 -0
- package/open-sse/handlers/fetch/index.js +237 -0
- package/open-sse/handlers/imageGenerationCore.js +230 -0
- package/open-sse/handlers/imageProviders/_base.js +31 -0
- package/open-sse/handlers/imageProviders/antigravity.js +73 -0
- package/open-sse/handlers/imageProviders/blackForestLabs.js +63 -0
- package/open-sse/handlers/imageProviders/cloudflareAi.js +207 -0
- package/open-sse/handlers/imageProviders/codex.js +217 -0
- package/open-sse/handlers/imageProviders/comfyui.js +25 -0
- package/open-sse/handlers/imageProviders/falAi.js +62 -0
- package/open-sse/handlers/imageProviders/gemini.js +40 -0
- package/open-sse/handlers/imageProviders/huggingface.js +38 -0
- package/open-sse/handlers/imageProviders/index.js +44 -0
- package/open-sse/handlers/imageProviders/leonardo.js +435 -0
- package/open-sse/handlers/imageProviders/nanobanana.js +79 -0
- package/open-sse/handlers/imageProviders/openai.js +61 -0
- package/open-sse/handlers/imageProviders/runwayml.js +72 -0
- package/open-sse/handlers/imageProviders/sdwebui.js +36 -0
- package/open-sse/handlers/imageProviders/stabilityAi.js +59 -0
- package/open-sse/handlers/imageProviders/weavy.js +222 -0
- package/open-sse/handlers/moderations.js +92 -0
- package/open-sse/handlers/ocr.js +89 -0
- package/open-sse/handlers/rerank.js +223 -0
- package/open-sse/handlers/responseSanitizer/reasoning.js +144 -0
- package/open-sse/handlers/responsesHandler.js +99 -0
- package/open-sse/handlers/search/callers.js +371 -0
- package/open-sse/handlers/search/chatSearch.js +409 -0
- package/open-sse/handlers/search/index.js +201 -0
- package/open-sse/handlers/search/normalizers.js +223 -0
- package/open-sse/handlers/search.js +1056 -0
- package/open-sse/handlers/sseParser.js +469 -0
- package/open-sse/handlers/sttCore.js +193 -0
- package/open-sse/handlers/ttsCore.js +74 -0
- package/open-sse/handlers/ttsProviders/_base.js +39 -0
- package/open-sse/handlers/ttsProviders/edgeTts.js +89 -0
- package/open-sse/handlers/ttsProviders/elevenlabs.js +48 -0
- package/open-sse/handlers/ttsProviders/gemini.js +117 -0
- package/open-sse/handlers/ttsProviders/genericFormats.js +169 -0
- package/open-sse/handlers/ttsProviders/googleTts.js +54 -0
- package/open-sse/handlers/ttsProviders/index.js +50 -0
- package/open-sse/handlers/ttsProviders/localDevice.js +87 -0
- package/open-sse/handlers/ttsProviders/minimax.js +59 -0
- package/open-sse/handlers/ttsProviders/openai.js +30 -0
- package/open-sse/handlers/ttsProviders/openrouter.js +70 -0
- package/open-sse/handlers/usageExtractor.js +93 -0
- package/open-sse/handlers/videoGenerationCore.js +119 -0
- package/open-sse/handlers/videoProviders/index.js +22 -0
- package/open-sse/handlers/videoProviders/leonardo.js +511 -0
- package/open-sse/handlers/videoProviders/weavy.js +198 -0
- package/open-sse/handlers/webFetch.js +200 -0
- package/open-sse/index.js +79 -0
- package/open-sse/mcp-server/audit.js +283 -0
- package/open-sse/mcp-server/catalog.js +184 -0
- package/open-sse/mcp-server/descriptionCompressor.js +175 -0
- package/open-sse/mcp-server/httpAuthContext.js +32 -0
- package/open-sse/mcp-server/httpTransport.js +229 -0
- package/open-sse/mcp-server/index.js +33 -0
- package/open-sse/mcp-server/mcpCallerIdentity.js +21 -0
- package/open-sse/mcp-server/runtimeHeartbeat.js +110 -0
- package/open-sse/mcp-server/schemas/a2a.js +153 -0
- package/open-sse/mcp-server/schemas/audit.js +17 -0
- package/open-sse/mcp-server/schemas/index.js +166 -0
- package/open-sse/mcp-server/schemas/pickFastestModel.js +90 -0
- package/open-sse/mcp-server/schemas/toolDefinition.js +0 -0
- package/open-sse/mcp-server/schemas/toolSearch.js +32 -0
- package/open-sse/mcp-server/schemas/tools.js +1219 -0
- package/open-sse/mcp-server/scopeEnforcement.js +80 -0
- package/open-sse/mcp-server/server.js +1168 -0
- package/open-sse/mcp-server/toolCardinality.js +78 -0
- package/open-sse/mcp-server/toolSearch/catalog.js +66 -0
- package/open-sse/mcp-server/toolSearch/handler.js +20 -0
- package/open-sse/mcp-server/toolSearch/index.js +10 -0
- package/open-sse/mcp-server/toolSearch/register.js +21 -0
- package/open-sse/mcp-server/toolSearch/search.js +55 -0
- package/open-sse/mcp-server/toolSearch/signature.js +67 -0
- package/open-sse/mcp-server/tools/advancedTools.js +893 -0
- package/open-sse/mcp-server/tools/agentSkillTools.js +69 -0
- package/open-sse/mcp-server/tools/compressionTools.js +309 -0
- package/open-sse/mcp-server/tools/gamificationTools.js +134 -0
- package/open-sse/mcp-server/tools/githubSkillTools.js +102 -0
- package/open-sse/mcp-server/tools/memoryTools.js +118 -0
- package/open-sse/mcp-server/tools/notionTools.js +101 -0
- package/open-sse/mcp-server/tools/obsidianTools.js +307 -0
- package/open-sse/mcp-server/tools/pickFastestModel.js +223 -0
- package/open-sse/mcp-server/tools/pluginTools.js +181 -0
- package/open-sse/mcp-server/tools/poolTools.js +139 -0
- package/open-sse/mcp-server/tools/skillTools.js +113 -0
- package/open-sse/package.json +10 -0
- package/open-sse/providers/REGISTRY_TEMPLATE.js +98 -0
- package/open-sse/providers/capabilities.js +293 -0
- package/open-sse/providers/index.js +51 -0
- package/open-sse/providers/models/helpers.js +20 -0
- package/open-sse/providers/models/namePatterns.js +33 -0
- package/open-sse/providers/models/schema.js +40 -0
- package/open-sse/providers/pricing.js +308 -0
- package/open-sse/providers/registry/alicode-intl.js +30 -0
- package/open-sse/providers/registry/alicode.js +31 -0
- package/open-sse/providers/registry/anthropic.js +37 -0
- package/open-sse/providers/registry/antigravity.js +81 -0
- package/open-sse/providers/registry/assemblyai.js +38 -0
- package/open-sse/providers/registry/aws-polly.js +45 -0
- package/open-sse/providers/registry/azure.js +21 -0
- package/open-sse/providers/registry/black-forest-labs.js +32 -0
- package/open-sse/providers/registry/blackbox.js +41 -0
- package/open-sse/providers/registry/brave-search.js +35 -0
- package/open-sse/providers/registry/byteplus.js +35 -0
- package/open-sse/providers/registry/cartesia.js +36 -0
- package/open-sse/providers/registry/cerebras.js +31 -0
- package/open-sse/providers/registry/chutes.js +24 -0
- package/open-sse/providers/registry/claude.js +87 -0
- package/open-sse/providers/registry/cline.js +51 -0
- package/open-sse/providers/registry/clinepass.js +57 -0
- package/open-sse/providers/registry/cloudflare-ai.js +55 -0
- package/open-sse/providers/registry/codebuddy-cn.js +77 -0
- package/open-sse/providers/registry/codex.js +100 -0
- package/open-sse/providers/registry/cohere.js +27 -0
- package/open-sse/providers/registry/comfyui.js +20 -0
- package/open-sse/providers/registry/commandcode.js +43 -0
- package/open-sse/providers/registry/coqui.js +30 -0
- package/open-sse/providers/registry/cursor.js +58 -0
- package/open-sse/providers/registry/deepgram.js +33 -0
- package/open-sse/providers/registry/deepseek.js +51 -0
- package/open-sse/providers/registry/digitalocean.js +92 -0
- package/open-sse/providers/registry/edge-tts.js +24 -0
- package/open-sse/providers/registry/elevenlabs.js +35 -0
- package/open-sse/providers/registry/exa.js +50 -0
- package/open-sse/providers/registry/fal-ai.js +34 -0
- package/open-sse/providers/registry/featherless.js +34 -0
- package/open-sse/providers/registry/firecrawl.js +34 -0
- package/open-sse/providers/registry/fireworks.js +29 -0
- package/open-sse/providers/registry/gemini-cli.js +58 -0
- package/open-sse/providers/registry/gemini.js +83 -0
- package/open-sse/providers/registry/github.js +88 -0
- package/open-sse/providers/registry/gitlab.js +32 -0
- package/open-sse/providers/registry/glm-cn.js +35 -0
- package/open-sse/providers/registry/glm.js +58 -0
- package/open-sse/providers/registry/google-pse.js +35 -0
- package/open-sse/providers/registry/google-tts.js +24 -0
- package/open-sse/providers/registry/grok-cli.js +86 -0
- package/open-sse/providers/registry/grok-web.js +39 -0
- package/open-sse/providers/registry/groq.js +37 -0
- package/open-sse/providers/registry/huggingface.js +34 -0
- package/open-sse/providers/registry/hyperbolic.js +35 -0
- package/open-sse/providers/registry/ibm-bob.js +27 -0
- package/open-sse/providers/registry/iflow.js +52 -0
- package/open-sse/providers/registry/index.js +210 -0
- package/open-sse/providers/registry/inworld.js +36 -0
- package/open-sse/providers/registry/jina-ai.js +42 -0
- package/open-sse/providers/registry/jina-reader.js +34 -0
- package/open-sse/providers/registry/kilocode.js +51 -0
- package/open-sse/providers/registry/kimchi.js +49 -0
- package/open-sse/providers/registry/kimi-coding.js +65 -0
- package/open-sse/providers/registry/kimi.js +56 -0
- package/open-sse/providers/registry/kiro.js +115 -0
- package/open-sse/providers/registry/linkup.js +34 -0
- package/open-sse/providers/registry/local-device.js +24 -0
- package/open-sse/providers/registry/mimo-free.js +24 -0
- package/open-sse/providers/registry/minimax-cn.js +76 -0
- package/open-sse/providers/registry/minimax.js +83 -0
- package/open-sse/providers/registry/mistral.js +34 -0
- package/open-sse/providers/registry/mmf.js +19 -0
- package/open-sse/providers/registry/nanobanana.js +35 -0
- package/open-sse/providers/registry/nebius.js +27 -0
- package/open-sse/providers/registry/nvidia.js +43 -0
- package/open-sse/providers/registry/ollama-local.js +19 -0
- package/open-sse/providers/registry/ollama.js +36 -0
- package/open-sse/providers/registry/openai.js +84 -0
- package/open-sse/providers/registry/opencode-go.js +41 -0
- package/open-sse/providers/registry/opencode.js +25 -0
- package/open-sse/providers/registry/openrouter.js +60 -0
- package/open-sse/providers/registry/perplexity-agent.js +49 -0
- package/open-sse/providers/registry/perplexity-web.js +33 -0
- package/open-sse/providers/registry/perplexity.js +37 -0
- package/open-sse/providers/registry/playht.js +36 -0
- package/open-sse/providers/registry/qoder.js +54 -0
- package/open-sse/providers/registry/qwen.js +33 -0
- package/open-sse/providers/registry/recraft.js +24 -0
- package/open-sse/providers/registry/runwayml.js +30 -0
- package/open-sse/providers/registry/sdwebui.js +20 -0
- package/open-sse/providers/registry/searchapi.js +35 -0
- package/open-sse/providers/registry/searxng.js +35 -0
- package/open-sse/providers/registry/serper.js +35 -0
- package/open-sse/providers/registry/siliconflow.js +39 -0
- package/open-sse/providers/registry/stability-ai.js +31 -0
- package/open-sse/providers/registry/tavily.js +50 -0
- package/open-sse/providers/registry/together.js +31 -0
- package/open-sse/providers/registry/topaz.js +19 -0
- package/open-sse/providers/registry/tortoise.js +30 -0
- package/open-sse/providers/registry/venice.js +56 -0
- package/open-sse/providers/registry/vercel-ai-gateway.js +41 -0
- package/open-sse/providers/registry/vertex-partner.js +29 -0
- package/open-sse/providers/registry/vertex.js +32 -0
- package/open-sse/providers/registry/volcengine-ark.js +35 -0
- package/open-sse/providers/registry/voyage-ai.js +30 -0
- package/open-sse/providers/registry/xai.js +45 -0
- package/open-sse/providers/registry/xiaomi-mimo.js +46 -0
- package/open-sse/providers/registry/xiaomi-tokenplan.js +63 -0
- package/open-sse/providers/registry/youcom.js +35 -0
- package/open-sse/providers/schema.js +76 -0
- package/open-sse/providers/shared.js +74 -0
- package/open-sse/providers/thinkingLevels.js +48 -0
- package/open-sse/rtk/applyFilter.js +15 -0
- package/open-sse/rtk/autodetect.js +111 -0
- package/open-sse/rtk/caveman.js +100 -0
- package/open-sse/rtk/cavemanPrompts.js +78 -0
- package/open-sse/rtk/constants.js +55 -0
- package/open-sse/rtk/filters/buildOutput.js +127 -0
- package/open-sse/rtk/filters/dedupLog.js +44 -0
- package/open-sse/rtk/filters/find.js +50 -0
- package/open-sse/rtk/filters/gitDiff.js +92 -0
- package/open-sse/rtk/filters/gitLog.js +99 -0
- package/open-sse/rtk/filters/gitStatus.js +117 -0
- package/open-sse/rtk/filters/grep.js +48 -0
- package/open-sse/rtk/filters/ls.js +79 -0
- package/open-sse/rtk/filters/readNumbered.js +27 -0
- package/open-sse/rtk/filters/searchList.js +52 -0
- package/open-sse/rtk/filters/smartTruncate.js +15 -0
- package/open-sse/rtk/filters/tree.js +32 -0
- package/open-sse/rtk/headroom.js +215 -0
- package/open-sse/rtk/index.js +155 -0
- package/open-sse/rtk/paleo.js +47 -0
- package/open-sse/rtk/ponytail.js +9 -0
- package/open-sse/rtk/ponytailPrompt.js +52 -0
- package/open-sse/rtk/ponytailPrompts.js +76 -0
- package/open-sse/rtk/pxpipe.js +104 -0
- package/open-sse/rtk/registry.js +38 -0
- package/open-sse/rtk/systemInject.js +98 -0
- package/open-sse/services/accountFallback.js +323 -0
- package/open-sse/services/antigravityHeaders.js +91 -0
- package/open-sse/services/antigravityVersion.js +154 -0
- package/open-sse/services/clinepassModels.js +63 -0
- package/open-sse/services/codexQuotaFetcher.js +312 -0
- package/open-sse/services/codexVerbosity.js +30 -0
- package/open-sse/services/combo/context.js +12 -0
- package/open-sse/services/combo/fingerprintExpansion.js +51 -0
- package/open-sse/services/combo/headroomRanking.js +28 -0
- package/open-sse/services/combo/quotaExhaustionCutoff.js +87 -0
- package/open-sse/services/combo/quotaScoring.js +226 -0
- package/open-sse/services/combo/quotaShareConcurrency.js +25 -0
- package/open-sse/services/combo/quotaShareInflight.js +50 -0
- package/open-sse/services/combo/quotaShareStrategy.js +130 -0
- package/open-sse/services/combo/resolveAutoStrategy.js +231 -0
- package/open-sse/services/combo/responseValidation.js +140 -0
- package/open-sse/services/combo/runtimeUnits.js +195 -0
- package/open-sse/services/combo/sessionStickiness.js +128 -0
- package/open-sse/services/combo/shadowRouting.js +131 -0
- package/open-sse/services/combo/targetExhaustion.js +75 -0
- package/open-sse/services/combo/targetSorters.js +126 -0
- package/open-sse/services/combo/targetTimeoutRunner.js +65 -0
- package/open-sse/services/combo/validateQuality.js +271 -0
- package/open-sse/services/combo.js +571 -0
- package/open-sse/services/comboAgentMiddleware.js +91 -0
- package/open-sse/services/comboConfig.js +153 -0
- package/open-sse/services/comboManifestMetrics.js +10 -0
- package/open-sse/services/comboMetrics.js +278 -0
- package/open-sse/services/compact.js +71 -0
- package/open-sse/services/copilotModels.js +155 -0
- package/open-sse/services/emergencyFallback.js +109 -0
- package/open-sse/services/errorClassifier.js +190 -0
- package/open-sse/services/evalRouting.js +201 -0
- package/open-sse/services/freeWebSearch.js +4 -0
- package/open-sse/services/fusion.js +223 -0
- package/open-sse/services/intentClassifier.js +631 -0
- package/open-sse/services/kimchiModels.js +176 -0
- package/open-sse/services/kiroModels.js +332 -0
- package/open-sse/services/model.js +136 -0
- package/open-sse/services/modelCapabilities.js +5 -0
- package/open-sse/services/modelDeprecation.js +101 -0
- package/open-sse/services/modelFamilyFallback.js +166 -0
- package/open-sse/services/oauthCredentialManager.js +156 -0
- package/open-sse/services/pipeline.js +107 -0
- package/open-sse/services/projectId.js +306 -0
- package/open-sse/services/provider.js +169 -0
- package/open-sse/services/providerCooldownTracker.js +94 -0
- package/open-sse/services/providerDefaultRateLimit.js +11 -0
- package/open-sse/services/proxyAutoSelector.js +24 -0
- package/open-sse/services/qoderModels.js +214 -0
- package/open-sse/services/quotaFetchThrottle.js +103 -0
- package/open-sse/services/quotaMonitor.js +310 -0
- package/open-sse/services/quotaPreflight.js +266 -0
- package/open-sse/services/rateLimitManager/headers.js +56 -0
- package/open-sse/services/rateLimitManager.js +616 -0
- package/open-sse/services/rateLimitSemaphore.js +112 -0
- package/open-sse/services/responsesInputSanitizer.js +99 -0
- package/open-sse/services/slidingWindowLimiter.js +44 -0
- package/open-sse/services/streamRecovery.js +355 -0
- package/open-sse/services/taskAwareRouter.js +227 -0
- package/open-sse/services/taskAwareRouting.js +332 -0
- package/open-sse/services/thinkingBudget.js +247 -0
- package/open-sse/services/tokenRefresh/dedup.js +31 -0
- package/open-sse/services/tokenRefresh/providers.js +624 -0
- package/open-sse/services/tokenRefresh.js +254 -0
- package/open-sse/services/usage/claude.js +147 -0
- package/open-sse/services/usage/codebuddy-cn.js +138 -0
- package/open-sse/services/usage/codex.js +267 -0
- package/open-sse/services/usage/github.js +100 -0
- package/open-sse/services/usage/google.js +241 -0
- package/open-sse/services/usage/grok-cli.js +274 -0
- package/open-sse/services/usage/kiro.js +188 -0
- package/open-sse/services/usage/minimax.js +234 -0
- package/open-sse/services/usage/misc.js +288 -0
- package/open-sse/services/usage/shared.js +70 -0
- package/open-sse/services/usage.js +62 -0
- package/open-sse/services/usageDb.js +2 -0
- package/open-sse/services/webSearchRouting.js +25 -0
- package/open-sse/services/wildcardRouter.js +59 -0
- package/open-sse/shared/clineAuth.js +37 -0
- package/open-sse/shared/machineId.js +19 -0
- package/open-sse/shared/qoder/constants.js +64 -0
- package/open-sse/shared/qoder/cosy.js +175 -0
- package/open-sse/shared/qoder/encoding.js +55 -0
- package/open-sse/transformer/responsesTransformer.js +439 -0
- package/open-sse/transformer/streamToJsonConverter.js +103 -0
- package/open-sse/translator/concerns/chunk.js +11 -0
- package/open-sse/translator/concerns/finishReason.js +63 -0
- package/open-sse/translator/concerns/image.js +124 -0
- package/open-sse/translator/concerns/json.js +5 -0
- package/open-sse/translator/concerns/message.js +7 -0
- package/open-sse/translator/concerns/modality.js +155 -0
- package/open-sse/translator/concerns/paramSupport.js +75 -0
- package/open-sse/translator/concerns/prefetch.js +96 -0
- package/open-sse/translator/concerns/reasoning.js +24 -0
- package/open-sse/translator/concerns/thinking.js +53 -0
- package/open-sse/translator/concerns/thinkingUnified.js +329 -0
- package/open-sse/translator/concerns/toolCall.js +153 -0
- package/open-sse/translator/concerns/usage.js +69 -0
- package/open-sse/translator/formats/claude.js +368 -0
- package/open-sse/translator/formats/gemini.js +380 -0
- package/open-sse/translator/formats/maxTokens.js +34 -0
- package/open-sse/translator/formats/openai.js +134 -0
- package/open-sse/translator/formats/responsesApi.js +141 -0
- package/open-sse/translator/formats.js +36 -0
- package/open-sse/translator/helpers/claudeHelper.js +216 -0
- package/open-sse/translator/helpers/geminiHelper.js +372 -0
- package/open-sse/translator/helpers/imageHelper.js +34 -0
- package/open-sse/translator/helpers/maxTokensHelper.js +27 -0
- package/open-sse/translator/helpers/openaiHelper.js +130 -0
- package/open-sse/translator/helpers/responsesApiHelper.js +139 -0
- package/open-sse/translator/helpers/toolCallHelper.js +148 -0
- package/open-sse/translator/index.js +252 -0
- package/open-sse/translator/request/antigravity-to-openai.js +242 -0
- package/open-sse/translator/request/claude-to-kiro.js +478 -0
- package/open-sse/translator/request/claude-to-openai.js +263 -0
- package/open-sse/translator/request/gemini-to-openai.js +152 -0
- package/open-sse/translator/request/openai-responses.js +385 -0
- package/open-sse/translator/request/openai-to-claude.js +383 -0
- package/open-sse/translator/request/openai-to-commandcode.js +176 -0
- package/open-sse/translator/request/openai-to-cursor.js +185 -0
- package/open-sse/translator/request/openai-to-gemini.js +442 -0
- package/open-sse/translator/request/openai-to-kiro.js +607 -0
- package/open-sse/translator/request/openai-to-kiro.old.js +278 -0
- package/open-sse/translator/request/openai-to-ollama.js +195 -0
- package/open-sse/translator/request/openai-to-vertex.js +42 -0
- package/open-sse/translator/response/claude-to-openai.js +195 -0
- package/open-sse/translator/response/commandcode-to-openai.js +188 -0
- package/open-sse/translator/response/cursor-to-openai.js +30 -0
- package/open-sse/translator/response/gemini-to-openai.js +147 -0
- package/open-sse/translator/response/kiro-to-claude.js +261 -0
- package/open-sse/translator/response/kiro-to-openai.js +160 -0
- package/open-sse/translator/response/ollama-to-openai.js +134 -0
- package/open-sse/translator/response/openai-responses.js +535 -0
- package/open-sse/translator/response/openai-to-antigravity.js +123 -0
- package/open-sse/translator/response/openai-to-claude.js +265 -0
- package/open-sse/translator/schema/blocks.js +43 -0
- package/open-sse/translator/schema/defaults.js +7 -0
- package/open-sse/translator/schema/finishReasons.js +27 -0
- package/open-sse/translator/schema/index.js +8 -0
- package/open-sse/translator/schema/roles.js +16 -0
- package/open-sse/utils/agentGoalPolicy.js +85 -0
- package/open-sse/utils/aiSdkCompat.js +75 -0
- package/open-sse/utils/bypassHandler.js +298 -0
- package/open-sse/utils/cacheControlPolicy.js +189 -0
- package/open-sse/utils/claudeCloaking.js +155 -0
- package/open-sse/utils/claudeHeaderCache.js +70 -0
- package/open-sse/utils/claudeSignature.js +41 -0
- package/open-sse/utils/clientDetector.js +63 -0
- package/open-sse/utils/codebuddySanitizer.js +96 -0
- package/open-sse/utils/composerToolCalls.js +158 -0
- package/open-sse/utils/cors.js +13 -0
- package/open-sse/utils/cursorChecksum.js +149 -0
- package/open-sse/utils/cursorImages.js +224 -0
- package/open-sse/utils/cursorProtobuf.js +904 -0
- package/open-sse/utils/cursorVersionDetector.js +55 -0
- package/open-sse/utils/debugLog.js +14 -0
- package/open-sse/utils/diagnostics.js +143 -0
- package/open-sse/utils/error.js +164 -0
- package/open-sse/utils/estimateSize.js +33 -0
- package/open-sse/utils/finishReason.js +32 -0
- package/open-sse/utils/flattenToolHistory.js +68 -0
- package/open-sse/utils/headers.js +36 -0
- package/open-sse/utils/heapPressure.js +35 -0
- package/open-sse/utils/jsonToSse.js +125 -0
- package/open-sse/utils/keepaliveThreshold.js +40 -0
- package/open-sse/utils/logger.js +104 -0
- package/open-sse/utils/noThinkingAlias.js +74 -0
- package/open-sse/utils/number.js +7 -0
- package/open-sse/utils/ollamaTransform.js +85 -0
- package/open-sse/utils/opencodeHeaders.js +47 -0
- package/open-sse/utils/progressTracker.js +81 -0
- package/open-sse/utils/providerRequestLogging.js +219 -0
- package/open-sse/utils/proxyFetch.js +369 -0
- package/open-sse/utils/reasoningContentInjector.js +79 -0
- package/open-sse/utils/reasoningFields.js +70 -0
- package/open-sse/utils/requestLogger.js +260 -0
- package/open-sse/utils/responsesInputNormalization.js +92 -0
- package/open-sse/utils/responsesStatePolicy.js +52 -0
- package/open-sse/utils/responsesStreamHelpers.js +49 -0
- package/open-sse/utils/sessionManager.js +240 -0
- package/open-sse/utils/sleep.js +6 -0
- package/open-sse/utils/sse.js +14 -0
- package/open-sse/utils/sseConstants.js +23 -0
- package/open-sse/utils/sseHeartbeat.js +104 -0
- package/open-sse/utils/stream.js +517 -0
- package/open-sse/utils/streamFailureFinalization.js +134 -0
- package/open-sse/utils/streamHandler.js +250 -0
- package/open-sse/utils/streamHelpers.js +137 -0
- package/open-sse/utils/textualToolCall.js +104 -0
- package/open-sse/utils/thinkCloseMarker.js +72 -0
- package/open-sse/utils/thinkTagParser.js +146 -0
- package/open-sse/utils/toolCallArguments.js +12 -0
- package/open-sse/utils/toolDeduper.js +49 -0
- package/open-sse/utils/toolSources.js +31 -0
- package/open-sse/utils/upstreamResponseHeaders.js +25 -0
- package/open-sse/utils/urlSanitize.js +15 -0
- package/open-sse/utils/usageTracking.js +347 -0
- package/package.json +1 -1
|
@@ -0,0 +1,226 @@
|
|
|
1
|
+
import { clamp01 } from "../../utils/number.ts";
|
|
2
|
+
import { isRecord } from "./comboData.ts";
|
|
3
|
+
import { RESET_WINDOW_NAMES } from "./types.ts";
|
|
4
|
+
const RESET_AWARE_SESSION_WINDOW_MS = 5 * 60 * 60 * 1e3;
|
|
5
|
+
const RESET_AWARE_WEEKLY_WINDOW_MS = 7 * 24 * 60 * 60 * 1e3;
|
|
6
|
+
const RESET_AWARE_SESSION_REMAINING_WEIGHT = 0.45;
|
|
7
|
+
const RESET_AWARE_SESSION_RESET_PRESSURE_WEIGHT = 0.55;
|
|
8
|
+
const RESET_AWARE_WEEKLY_REMAINING_WEIGHT = 0.25;
|
|
9
|
+
const RESET_AWARE_WEEKLY_RESET_PRESSURE_WEIGHT = 0.75;
|
|
10
|
+
const RESET_AWARE_DEFAULTS = {
|
|
11
|
+
sessionWeight: 0.35,
|
|
12
|
+
weeklyWeight: 0.65,
|
|
13
|
+
tieBandPercent: 5,
|
|
14
|
+
exhaustionGuardPercent: 10
|
|
15
|
+
};
|
|
16
|
+
const RESET_WINDOW_DEFAULT_TIE_BAND_MS = 6e4;
|
|
17
|
+
function finiteNumberOrNull(value) {
|
|
18
|
+
const numericValue = Number(value);
|
|
19
|
+
return Number.isFinite(numericValue) ? numericValue : null;
|
|
20
|
+
}
|
|
21
|
+
function getPercentConfig(value, fallback) {
|
|
22
|
+
const numericValue = finiteNumberOrNull(value);
|
|
23
|
+
if (numericValue === null) return fallback;
|
|
24
|
+
return Math.max(0, Math.min(100, numericValue));
|
|
25
|
+
}
|
|
26
|
+
function getWeightConfig(value, fallback) {
|
|
27
|
+
const numericValue = finiteNumberOrNull(value);
|
|
28
|
+
if (numericValue === null || numericValue < 0) return fallback;
|
|
29
|
+
return numericValue;
|
|
30
|
+
}
|
|
31
|
+
function getDurationConfig(value, fallback, max) {
|
|
32
|
+
const numericValue = finiteNumberOrNull(value);
|
|
33
|
+
if (numericValue === null || numericValue < 0) return fallback;
|
|
34
|
+
return Math.min(max, Math.floor(numericValue));
|
|
35
|
+
}
|
|
36
|
+
function resolveResetAwareConfig(config) {
|
|
37
|
+
const sessionWeight = getWeightConfig(
|
|
38
|
+
config?.resetAwareSessionWeight,
|
|
39
|
+
RESET_AWARE_DEFAULTS.sessionWeight
|
|
40
|
+
);
|
|
41
|
+
const weeklyWeight = getWeightConfig(
|
|
42
|
+
config?.resetAwareWeeklyWeight,
|
|
43
|
+
RESET_AWARE_DEFAULTS.weeklyWeight
|
|
44
|
+
);
|
|
45
|
+
const totalWeight = sessionWeight + weeklyWeight;
|
|
46
|
+
const normalizedSessionWeight = totalWeight > 0 ? sessionWeight / totalWeight : RESET_AWARE_DEFAULTS.sessionWeight;
|
|
47
|
+
return {
|
|
48
|
+
sessionWeight: normalizedSessionWeight,
|
|
49
|
+
weeklyWeight: 1 - normalizedSessionWeight,
|
|
50
|
+
tieBand: getPercentConfig(config?.resetAwareTieBandPercent, RESET_AWARE_DEFAULTS.tieBandPercent) / 100,
|
|
51
|
+
exhaustionGuard: getPercentConfig(
|
|
52
|
+
config?.resetAwareExhaustionGuardPercent,
|
|
53
|
+
RESET_AWARE_DEFAULTS.exhaustionGuardPercent
|
|
54
|
+
) / 100,
|
|
55
|
+
quotaCacheTtlMs: getDurationConfig(config?.resetAwareQuotaCacheTtlMs, 0, 3e5),
|
|
56
|
+
quotaCacheMaxStaleMs: getDurationConfig(config?.resetAwareQuotaCacheMaxStaleMs, 0, 36e5)
|
|
57
|
+
};
|
|
58
|
+
}
|
|
59
|
+
function resolveResetWindowConfig(config) {
|
|
60
|
+
const rawWindows = Array.isArray(config?.resetWindowWindows) ? config.resetWindowWindows : null;
|
|
61
|
+
const windows = rawWindows?.filter(
|
|
62
|
+
(windowName) => RESET_WINDOW_NAMES.includes(String(windowName))
|
|
63
|
+
).filter((windowName, index, array) => array.indexOf(windowName) === index);
|
|
64
|
+
const effectiveWindows = windows && windows.length > 0 ? windows : config?.resetWindowIncludeSession === true ? ["weekly", "session"] : ["weekly"];
|
|
65
|
+
return {
|
|
66
|
+
windows: effectiveWindows,
|
|
67
|
+
tieBandMs: Math.max(
|
|
68
|
+
0,
|
|
69
|
+
finiteNumberOrNull(config?.resetWindowTieBandMs) ?? RESET_WINDOW_DEFAULT_TIE_BAND_MS
|
|
70
|
+
),
|
|
71
|
+
quotaCacheTtlMs: getDurationConfig(config?.resetWindowQuotaCacheTtlMs, 0, 3e5),
|
|
72
|
+
quotaCacheMaxStaleMs: getDurationConfig(config?.resetWindowQuotaCacheMaxStaleMs, 0, 36e5)
|
|
73
|
+
};
|
|
74
|
+
}
|
|
75
|
+
function resolveSlaRoutingPolicy(config) {
|
|
76
|
+
if (!config) return void 0;
|
|
77
|
+
const nestedSla = isRecord(config.sla) ? config.sla : {};
|
|
78
|
+
const targetP95Ms = finiteNumberOrNull(config.slaTargetP95Ms ?? nestedSla.targetP95Ms);
|
|
79
|
+
const maxErrorRate = finiteNumberOrNull(config.slaMaxErrorRate ?? nestedSla.maxErrorRate);
|
|
80
|
+
const maxCostPer1MTokens = finiteNumberOrNull(
|
|
81
|
+
config.slaMaxCostPer1MTokens ?? nestedSla.maxCostPer1MTokens
|
|
82
|
+
);
|
|
83
|
+
const hardConstraints = config.slaHardConstraints ?? nestedSla.hardConstraints;
|
|
84
|
+
const policy = {};
|
|
85
|
+
if (targetP95Ms !== null && targetP95Ms > 0) policy.targetP95Ms = targetP95Ms;
|
|
86
|
+
if (maxErrorRate !== null && maxErrorRate >= 0) policy.maxErrorRate = clamp01(maxErrorRate);
|
|
87
|
+
if (maxCostPer1MTokens !== null && maxCostPer1MTokens > 0) {
|
|
88
|
+
policy.maxCostPer1MTokens = maxCostPer1MTokens;
|
|
89
|
+
}
|
|
90
|
+
if (typeof hardConstraints === "boolean") policy.hardConstraints = hardConstraints;
|
|
91
|
+
return Object.keys(policy).length > 0 ? policy : void 0;
|
|
92
|
+
}
|
|
93
|
+
function getResetAwareProvider(target) {
|
|
94
|
+
const provider = (target.providerId || target.provider || "").toLowerCase();
|
|
95
|
+
return provider || null;
|
|
96
|
+
}
|
|
97
|
+
function normalizeResetAt(value) {
|
|
98
|
+
if (typeof value === "string" && value.trim().length > 0) return value.trim();
|
|
99
|
+
if (typeof value === "number" && Number.isFinite(value)) return String(value);
|
|
100
|
+
return null;
|
|
101
|
+
}
|
|
102
|
+
function parseResetTimeMs(resetAt) {
|
|
103
|
+
if (!resetAt) return NaN;
|
|
104
|
+
const resetTime = Date.parse(resetAt);
|
|
105
|
+
if (Number.isFinite(resetTime)) return resetTime;
|
|
106
|
+
if (!/^\d+(?:\.\d+)?$/.test(resetAt)) return NaN;
|
|
107
|
+
const numericResetAt = Number(resetAt);
|
|
108
|
+
if (!Number.isFinite(numericResetAt)) return NaN;
|
|
109
|
+
return numericResetAt < 1e10 ? numericResetAt * 1e3 : numericResetAt;
|
|
110
|
+
}
|
|
111
|
+
function getQuotaWindow(quota, key) {
|
|
112
|
+
if (!isRecord(quota)) return null;
|
|
113
|
+
const window = quota[key];
|
|
114
|
+
if (!isRecord(window)) return null;
|
|
115
|
+
const percentUsed = finiteNumberOrNull(window.percentUsed);
|
|
116
|
+
const resetAt = normalizeResetAt(window.resetAt);
|
|
117
|
+
return { percentUsed, resetAt };
|
|
118
|
+
}
|
|
119
|
+
function normalizeWindowPercentUsed(value) {
|
|
120
|
+
const numericValue = finiteNumberOrNull(value);
|
|
121
|
+
if (numericValue === null) return null;
|
|
122
|
+
if (numericValue > 1) return clamp01(numericValue / 100);
|
|
123
|
+
return clamp01(numericValue);
|
|
124
|
+
}
|
|
125
|
+
function getNamedQuotaWindow(quota, windowName) {
|
|
126
|
+
if (!quota || !isRecord(quota)) return null;
|
|
127
|
+
if (windowName === "session") return getQuotaWindow(quota, "window5h");
|
|
128
|
+
if (windowName === "weekly") {
|
|
129
|
+
return getQuotaWindow(quota, "window7d") || getQuotaWindow(quota, "windowWeekly");
|
|
130
|
+
}
|
|
131
|
+
if (windowName === "monthly") return getQuotaWindow(quota, "windowMonthly");
|
|
132
|
+
return null;
|
|
133
|
+
}
|
|
134
|
+
function getWindowsMapQuotaWindow(quota, windowName) {
|
|
135
|
+
if (!quota || !isRecord(quota) || !isRecord(quota.windows)) return null;
|
|
136
|
+
const candidates = Object.entries(quota.windows).map(([key, value]) => ({ key: key.toLowerCase(), value })).filter(({ key }) => key === windowName || key.startsWith(`${windowName} `));
|
|
137
|
+
if (candidates.length === 0) return null;
|
|
138
|
+
candidates.sort((a, b) => a.key.localeCompare(b.key));
|
|
139
|
+
const window = candidates[0].value;
|
|
140
|
+
if (!isRecord(window)) return null;
|
|
141
|
+
return {
|
|
142
|
+
percentUsed: normalizeWindowPercentUsed(window.percentUsed),
|
|
143
|
+
resetAt: normalizeResetAt(window.resetAt)
|
|
144
|
+
};
|
|
145
|
+
}
|
|
146
|
+
function resolveQuotaWindowByName(quota, windowName) {
|
|
147
|
+
return getNamedQuotaWindow(quota, windowName) || getWindowsMapQuotaWindow(quota, windowName);
|
|
148
|
+
}
|
|
149
|
+
function getResetUrgency(resetAt, windowMs) {
|
|
150
|
+
if (!resetAt) return 0.5;
|
|
151
|
+
const resetTime = parseResetTimeMs(resetAt);
|
|
152
|
+
if (!Number.isFinite(resetTime)) return 0.5;
|
|
153
|
+
const msUntilReset = resetTime - Date.now();
|
|
154
|
+
if (msUntilReset <= 0) return 1;
|
|
155
|
+
return clamp01(1 - msUntilReset / windowMs);
|
|
156
|
+
}
|
|
157
|
+
function scoreQuotaWindow(remaining, resetAt, windowMs, remainingWeight, resetPressureWeight) {
|
|
158
|
+
const normalizedRemaining = clamp01(remaining);
|
|
159
|
+
const resetUrgency = getResetUrgency(resetAt, windowMs);
|
|
160
|
+
const resetPressure = resetUrgency * (1 - normalizedRemaining);
|
|
161
|
+
return remainingWeight * normalizedRemaining + resetPressureWeight * resetPressure;
|
|
162
|
+
}
|
|
163
|
+
function scoreResetAwareQuota(quota, config) {
|
|
164
|
+
if (!quota || !isRecord(quota)) return { score: 0.5 };
|
|
165
|
+
if (quota.limitReached === true) return { score: -Infinity };
|
|
166
|
+
const overallPercentUsed = clamp01(finiteNumberOrNull(quota.percentUsed) ?? 0.5);
|
|
167
|
+
const sessionWindow = getQuotaWindow(quota, "window5h");
|
|
168
|
+
const weeklyWindow = getQuotaWindow(quota, "window7d") || getQuotaWindow(quota, "windowWeekly");
|
|
169
|
+
const sessionRemaining = clamp01(1 - (sessionWindow?.percentUsed ?? overallPercentUsed));
|
|
170
|
+
const weeklyRemaining = clamp01(1 - (weeklyWindow?.percentUsed ?? overallPercentUsed));
|
|
171
|
+
const sessionScore = scoreQuotaWindow(
|
|
172
|
+
sessionRemaining,
|
|
173
|
+
sessionWindow?.resetAt,
|
|
174
|
+
RESET_AWARE_SESSION_WINDOW_MS,
|
|
175
|
+
RESET_AWARE_SESSION_REMAINING_WEIGHT,
|
|
176
|
+
RESET_AWARE_SESSION_RESET_PRESSURE_WEIGHT
|
|
177
|
+
);
|
|
178
|
+
const weeklyScore = scoreQuotaWindow(
|
|
179
|
+
weeklyRemaining,
|
|
180
|
+
weeklyWindow?.resetAt ?? normalizeResetAt(quota.resetAt),
|
|
181
|
+
RESET_AWARE_WEEKLY_WINDOW_MS,
|
|
182
|
+
RESET_AWARE_WEEKLY_REMAINING_WEIGHT,
|
|
183
|
+
RESET_AWARE_WEEKLY_RESET_PRESSURE_WEIGHT
|
|
184
|
+
);
|
|
185
|
+
let score = config.sessionWeight * sessionScore + config.weeklyWeight * weeklyScore;
|
|
186
|
+
if (config.exhaustionGuard > 0 && sessionRemaining < config.exhaustionGuard) {
|
|
187
|
+
score *= Math.max(0.05, sessionRemaining / config.exhaustionGuard);
|
|
188
|
+
}
|
|
189
|
+
return { score };
|
|
190
|
+
}
|
|
191
|
+
function getResetWindowTimestampMs(quota, windows) {
|
|
192
|
+
if (!quota || !isRecord(quota) || quota.limitReached === true) return Infinity;
|
|
193
|
+
let selectedResetMs = Infinity;
|
|
194
|
+
for (const windowName of windows) {
|
|
195
|
+
const window = resolveQuotaWindowByName(quota, windowName);
|
|
196
|
+
const resetMs = parseResetTimeMs(window?.resetAt ?? null);
|
|
197
|
+
if (Number.isFinite(resetMs)) {
|
|
198
|
+
selectedResetMs = Math.min(selectedResetMs, resetMs);
|
|
199
|
+
}
|
|
200
|
+
}
|
|
201
|
+
if (!Number.isFinite(selectedResetMs)) {
|
|
202
|
+
selectedResetMs = parseResetTimeMs(normalizeResetAt(quota.resetAt));
|
|
203
|
+
}
|
|
204
|
+
return Number.isFinite(selectedResetMs) ? selectedResetMs : Infinity;
|
|
205
|
+
}
|
|
206
|
+
function getResetWindowHorizonMs(windows) {
|
|
207
|
+
if (windows.includes("monthly")) return 30 * 24 * 60 * 60 * 1e3;
|
|
208
|
+
if (windows.includes("weekly")) return RESET_AWARE_WEEKLY_WINDOW_MS;
|
|
209
|
+
return RESET_AWARE_SESSION_WINDOW_MS;
|
|
210
|
+
}
|
|
211
|
+
function calculateResetWindowAffinity(quota, config) {
|
|
212
|
+
const resetMs = getResetWindowTimestampMs(quota, config.windows);
|
|
213
|
+
if (!Number.isFinite(resetMs)) return 0.5;
|
|
214
|
+
const msUntilReset = resetMs - Date.now();
|
|
215
|
+
if (msUntilReset <= 0) return 1;
|
|
216
|
+
return clamp01(1 - msUntilReset / getResetWindowHorizonMs(config.windows));
|
|
217
|
+
}
|
|
218
|
+
export {
|
|
219
|
+
calculateResetWindowAffinity,
|
|
220
|
+
getResetAwareProvider,
|
|
221
|
+
getResetWindowTimestampMs,
|
|
222
|
+
resolveResetAwareConfig,
|
|
223
|
+
resolveResetWindowConfig,
|
|
224
|
+
resolveSlaRoutingPolicy,
|
|
225
|
+
scoreResetAwareQuota
|
|
226
|
+
};
|
|
@@ -0,0 +1,25 @@
|
|
|
1
|
+
import * as semaphore from "../rateLimitSemaphore.ts";
|
|
2
|
+
function quotaShareConcurrencyKey(connectionId) {
|
|
3
|
+
return `qsconn:${connectionId}`;
|
|
4
|
+
}
|
|
5
|
+
async function acquireQuotaShareConcurrencySlot(target, cap, opts, log) {
|
|
6
|
+
const connectionId = target?.connectionId ?? "";
|
|
7
|
+
if (!connectionId || cap === null || cap <= 0) return null;
|
|
8
|
+
try {
|
|
9
|
+
return await semaphore.acquire(quotaShareConcurrencyKey(connectionId), {
|
|
10
|
+
maxConcurrency: cap,
|
|
11
|
+
timeoutMs: opts.queueTimeoutMs,
|
|
12
|
+
maxQueueSize: opts.maxQueueSize
|
|
13
|
+
});
|
|
14
|
+
} catch {
|
|
15
|
+
log.warn(
|
|
16
|
+
"COMBO",
|
|
17
|
+
`Quota-share concurrency: connection ${connectionId} gate saturated (cap=${cap}) \u2014 proceeding without a slot`
|
|
18
|
+
);
|
|
19
|
+
return null;
|
|
20
|
+
}
|
|
21
|
+
}
|
|
22
|
+
export {
|
|
23
|
+
acquireQuotaShareConcurrencySlot,
|
|
24
|
+
quotaShareConcurrencyKey
|
|
25
|
+
};
|
|
@@ -0,0 +1,50 @@
|
|
|
1
|
+
const DEFAULT_LEASE_MS = 12e4;
|
|
2
|
+
const _inflightMap = /* @__PURE__ */ new Map();
|
|
3
|
+
function incrementInflight(connectionId, leaseMs = DEFAULT_LEASE_MS, nowMs = Date.now()) {
|
|
4
|
+
if (!connectionId) return 0;
|
|
5
|
+
pruneExpired(nowMs);
|
|
6
|
+
const slot = _inflightMap.get(connectionId);
|
|
7
|
+
const base = slot && slot.expiresAtMs > nowMs ? slot.count : 0;
|
|
8
|
+
const newCount = base + 1;
|
|
9
|
+
_inflightMap.set(connectionId, { count: newCount, expiresAtMs: nowMs + leaseMs });
|
|
10
|
+
return newCount;
|
|
11
|
+
}
|
|
12
|
+
function decrementInflight(connectionId, nowMs = Date.now()) {
|
|
13
|
+
if (!connectionId) return;
|
|
14
|
+
const slot = _inflightMap.get(connectionId);
|
|
15
|
+
if (!slot || slot.expiresAtMs <= nowMs) {
|
|
16
|
+
_inflightMap.delete(connectionId);
|
|
17
|
+
return;
|
|
18
|
+
}
|
|
19
|
+
const newCount = Math.max(0, slot.count - 1);
|
|
20
|
+
if (newCount === 0) {
|
|
21
|
+
_inflightMap.delete(connectionId);
|
|
22
|
+
} else {
|
|
23
|
+
_inflightMap.set(connectionId, { count: newCount, expiresAtMs: slot.expiresAtMs });
|
|
24
|
+
}
|
|
25
|
+
}
|
|
26
|
+
function getInflight(connectionId, nowMs = Date.now()) {
|
|
27
|
+
if (!connectionId) return 0;
|
|
28
|
+
const slot = _inflightMap.get(connectionId);
|
|
29
|
+
if (!slot || slot.expiresAtMs <= nowMs) return 0;
|
|
30
|
+
return slot.count;
|
|
31
|
+
}
|
|
32
|
+
function pruneExpired(nowMs) {
|
|
33
|
+
for (const [key, slot] of _inflightMap) {
|
|
34
|
+
if (slot.expiresAtMs <= nowMs) _inflightMap.delete(key);
|
|
35
|
+
}
|
|
36
|
+
}
|
|
37
|
+
function _clearInflightForTest() {
|
|
38
|
+
_inflightMap.clear();
|
|
39
|
+
}
|
|
40
|
+
function _inflightSizeForTest() {
|
|
41
|
+
return _inflightMap.size;
|
|
42
|
+
}
|
|
43
|
+
export {
|
|
44
|
+
DEFAULT_LEASE_MS,
|
|
45
|
+
_clearInflightForTest,
|
|
46
|
+
_inflightSizeForTest,
|
|
47
|
+
decrementInflight,
|
|
48
|
+
getInflight,
|
|
49
|
+
incrementInflight
|
|
50
|
+
};
|
|
@@ -0,0 +1,130 @@
|
|
|
1
|
+
import { isBucketSaturated } from '../../../dist/lib/quota/accountBuckets.ts.js';
|
|
2
|
+
import { incrementInflight, decrementInflight, getInflight } from "./quotaShareInflight.ts";
|
|
3
|
+
const MAX_DRR_COMBOS = 200;
|
|
4
|
+
const _drrState = /* @__PURE__ */ new Map();
|
|
5
|
+
function getDrrDeficits(comboName) {
|
|
6
|
+
let deficits = _drrState.get(comboName);
|
|
7
|
+
if (!deficits) {
|
|
8
|
+
if (_drrState.size >= MAX_DRR_COMBOS) {
|
|
9
|
+
const oldestKey = _drrState.keys().next().value;
|
|
10
|
+
if (oldestKey !== void 0) _drrState.delete(oldestKey);
|
|
11
|
+
}
|
|
12
|
+
deficits = /* @__PURE__ */ new Map();
|
|
13
|
+
_drrState.set(comboName, deficits);
|
|
14
|
+
}
|
|
15
|
+
return deficits;
|
|
16
|
+
}
|
|
17
|
+
function bareModelName(modelStr) {
|
|
18
|
+
const slash = modelStr.indexOf("/");
|
|
19
|
+
return slash >= 0 ? modelStr.slice(slash + 1) : modelStr;
|
|
20
|
+
}
|
|
21
|
+
function filterEligibleBySaturation(targets, modelStr, nowMs) {
|
|
22
|
+
const modelName = bareModelName(modelStr);
|
|
23
|
+
const eligible = targets.filter((target) => {
|
|
24
|
+
const connId = target.connectionId ?? "";
|
|
25
|
+
if (connId === "") return true;
|
|
26
|
+
const saturated = isBucketSaturated(connId, "5h", nowMs) || isBucketSaturated(connId, "7d", nowMs) || modelName !== "" && isBucketSaturated(connId, `7d:${modelName}`, nowMs);
|
|
27
|
+
return !saturated;
|
|
28
|
+
});
|
|
29
|
+
return eligible.length > 0 ? eligible : targets;
|
|
30
|
+
}
|
|
31
|
+
function resolveConnectionCap(connectionId, caps) {
|
|
32
|
+
if (!connectionId || !caps) return null;
|
|
33
|
+
const cap = caps.get(connectionId);
|
|
34
|
+
if (typeof cap !== "number" || !Number.isFinite(cap) || cap <= 0) return null;
|
|
35
|
+
return cap;
|
|
36
|
+
}
|
|
37
|
+
function partitionByConcurrencyCap(targets, caps, nowMs) {
|
|
38
|
+
if (!caps || caps.size === 0) return { withRoom: targets, atCap: [] };
|
|
39
|
+
const withRoom = [];
|
|
40
|
+
const atCap = [];
|
|
41
|
+
for (const target of targets) {
|
|
42
|
+
const connId = target.connectionId ?? "";
|
|
43
|
+
const cap = resolveConnectionCap(connId, caps);
|
|
44
|
+
if (cap === null) {
|
|
45
|
+
withRoom.push(target);
|
|
46
|
+
continue;
|
|
47
|
+
}
|
|
48
|
+
if (getInflight(connId, nowMs) >= cap) {
|
|
49
|
+
atCap.push(target);
|
|
50
|
+
} else {
|
|
51
|
+
withRoom.push(target);
|
|
52
|
+
}
|
|
53
|
+
}
|
|
54
|
+
if (withRoom.length === 0) return { withRoom: targets, atCap: [] };
|
|
55
|
+
return { withRoom, atCap };
|
|
56
|
+
}
|
|
57
|
+
function applyDrr(targets, comboName) {
|
|
58
|
+
if (targets.length <= 1) return targets.slice();
|
|
59
|
+
const deficits = getDrrDeficits(comboName);
|
|
60
|
+
const totalWeight = targets.reduce((sum, t) => sum + normalizeWeight(t.weight), 0);
|
|
61
|
+
for (const target of targets) {
|
|
62
|
+
const quantum = normalizeWeight(target.weight) / totalWeight;
|
|
63
|
+
deficits.set(target.executionKey, (deficits.get(target.executionKey) ?? 0) + quantum);
|
|
64
|
+
}
|
|
65
|
+
let winner = targets[0];
|
|
66
|
+
let bestDeficit = deficits.get(winner.executionKey) ?? 0;
|
|
67
|
+
for (let i = 1; i < targets.length; i++) {
|
|
68
|
+
const d = deficits.get(targets[i].executionKey) ?? 0;
|
|
69
|
+
if (d > bestDeficit) {
|
|
70
|
+
bestDeficit = d;
|
|
71
|
+
winner = targets[i];
|
|
72
|
+
}
|
|
73
|
+
}
|
|
74
|
+
deficits.set(winner.executionKey, bestDeficit - 1);
|
|
75
|
+
const rest = targets.filter((t) => t.executionKey !== winner.executionKey);
|
|
76
|
+
return [winner, ...rest];
|
|
77
|
+
}
|
|
78
|
+
function normalizeWeight(weight) {
|
|
79
|
+
return Number.isFinite(weight) && weight > 0 ? weight : 1;
|
|
80
|
+
}
|
|
81
|
+
function pickByInflightP2C(first, second, nowMs) {
|
|
82
|
+
const loadFirst = getInflight(first.connectionId ?? "", nowMs);
|
|
83
|
+
const loadSecond = getInflight(second.connectionId ?? "", nowMs);
|
|
84
|
+
return loadSecond < loadFirst ? 1 : 0;
|
|
85
|
+
}
|
|
86
|
+
function selectQuotaShareTarget(targets, comboName, modelStr, nowMs = Date.now(), options) {
|
|
87
|
+
const noOp = () => {
|
|
88
|
+
};
|
|
89
|
+
if (targets.length === 0) {
|
|
90
|
+
return { target: null, orderedTargets: [], decrementInflight: noOp };
|
|
91
|
+
}
|
|
92
|
+
const eligible = filterEligibleBySaturation(targets, modelStr, nowMs);
|
|
93
|
+
const saturatedDeprioritized = targets.filter((t) => !eligible.includes(t));
|
|
94
|
+
const { withRoom, atCap } = partitionByConcurrencyCap(
|
|
95
|
+
eligible,
|
|
96
|
+
options?.maxConcurrentByConnection,
|
|
97
|
+
nowMs
|
|
98
|
+
);
|
|
99
|
+
const ordered = applyDrr(withRoom, comboName);
|
|
100
|
+
let winner;
|
|
101
|
+
let rest;
|
|
102
|
+
if (ordered.length >= 2 && pickByInflightP2C(ordered[0], ordered[1], nowMs) === 1) {
|
|
103
|
+
winner = ordered[1];
|
|
104
|
+
rest = [ordered[0], ...ordered.slice(2)];
|
|
105
|
+
} else {
|
|
106
|
+
winner = ordered[0];
|
|
107
|
+
rest = ordered.slice(1);
|
|
108
|
+
}
|
|
109
|
+
const winnerConnectionId = winner.connectionId ?? "";
|
|
110
|
+
if (winnerConnectionId) incrementInflight(winnerConnectionId, void 0, nowMs);
|
|
111
|
+
const orderedTargets = [winner, ...rest, ...atCap, ...saturatedDeprioritized];
|
|
112
|
+
let released = false;
|
|
113
|
+
const release = () => {
|
|
114
|
+
if (released) return;
|
|
115
|
+
released = true;
|
|
116
|
+
if (winnerConnectionId) decrementInflight(winnerConnectionId, nowMs);
|
|
117
|
+
};
|
|
118
|
+
return { target: winner, orderedTargets, decrementInflight: release };
|
|
119
|
+
}
|
|
120
|
+
function _clearDrrStateForTest() {
|
|
121
|
+
_drrState.clear();
|
|
122
|
+
}
|
|
123
|
+
function _getDrrDeficitForTest(comboName, executionKey) {
|
|
124
|
+
return _drrState.get(comboName)?.get(executionKey) ?? 0;
|
|
125
|
+
}
|
|
126
|
+
export {
|
|
127
|
+
_clearDrrStateForTest,
|
|
128
|
+
_getDrrDeficitForTest,
|
|
129
|
+
selectQuotaShareTarget
|
|
130
|
+
};
|
|
@@ -0,0 +1,231 @@
|
|
|
1
|
+
import { unavailableResponse } from "../../utils/error.ts";
|
|
2
|
+
import { selectProvider as selectAutoProvider } from "../autoCombo/engine.ts";
|
|
3
|
+
import {
|
|
4
|
+
resolveRequestModePack,
|
|
5
|
+
parseRequestBudgetCap
|
|
6
|
+
} from "../autoCombo/requestControls.ts";
|
|
7
|
+
import { selectWithStrategy } from "../autoCombo/routerStrategy.ts";
|
|
8
|
+
import { buildComplexityRoutingHint } from "../autoCombo/complexityRouter";
|
|
9
|
+
import { recordComboIntent } from "../comboMetrics.ts";
|
|
10
|
+
import { estimateTokens } from "../contextManager.ts";
|
|
11
|
+
import { classifyWithConfig } from "../intentClassifier.ts";
|
|
12
|
+
import { parseModel } from "../model.ts";
|
|
13
|
+
import { supportsToolCalling } from "../modelCapabilities.ts";
|
|
14
|
+
import { parseAutoConfig } from "./autoConfig.ts";
|
|
15
|
+
import { dedupeTargetsByExecutionKey } from "./comboData.ts";
|
|
16
|
+
import { getModelContextLimitForModelString } from "./comboStructure.ts";
|
|
17
|
+
import {
|
|
18
|
+
_registerExecutionCandidates,
|
|
19
|
+
expandAutoComboCandidatePool,
|
|
20
|
+
extractPromptForIntent,
|
|
21
|
+
getIntentConfig,
|
|
22
|
+
mapIntentToTaskType,
|
|
23
|
+
scoreAutoTargets
|
|
24
|
+
} from "./autoStrategy.ts";
|
|
25
|
+
async function resolveAutoStrategyOrder(deps) {
|
|
26
|
+
const {
|
|
27
|
+
body,
|
|
28
|
+
combo,
|
|
29
|
+
settings,
|
|
30
|
+
config,
|
|
31
|
+
relayOptions,
|
|
32
|
+
resilienceSettings,
|
|
33
|
+
log,
|
|
34
|
+
buildAutoCandidates
|
|
35
|
+
} = deps;
|
|
36
|
+
let orderedTargets = deps.orderedTargets;
|
|
37
|
+
let autoUsedExplicitRouter = false;
|
|
38
|
+
const requestHasTools = Array.isArray(body?.tools) && body.tools.length > 0;
|
|
39
|
+
let eligibleTargets = [...orderedTargets];
|
|
40
|
+
if (requestHasTools) {
|
|
41
|
+
const filtered = eligibleTargets.filter((target) => supportsToolCalling(target.modelStr));
|
|
42
|
+
if (filtered.length > 0) {
|
|
43
|
+
eligibleTargets = filtered;
|
|
44
|
+
} else {
|
|
45
|
+
log.warn(
|
|
46
|
+
"COMBO",
|
|
47
|
+
"Auto strategy: all candidates filtered by tool-calling policy, falling back to full pool"
|
|
48
|
+
);
|
|
49
|
+
}
|
|
50
|
+
}
|
|
51
|
+
const requestMessages = body.messages;
|
|
52
|
+
const estimatedInputTokens = estimateTokens(
|
|
53
|
+
typeof requestMessages === "string" || requestMessages !== null && typeof requestMessages === "object" ? requestMessages : []
|
|
54
|
+
);
|
|
55
|
+
if (estimatedInputTokens > 0) {
|
|
56
|
+
const filteredByContext = eligibleTargets.filter((target) => {
|
|
57
|
+
const limit = getModelContextLimitForModelString(target.modelStr);
|
|
58
|
+
if (limit === null || limit === void 0) return true;
|
|
59
|
+
return limit >= estimatedInputTokens;
|
|
60
|
+
});
|
|
61
|
+
if (filteredByContext.length > 0) {
|
|
62
|
+
log.debug?.(
|
|
63
|
+
"COMBO",
|
|
64
|
+
`Auto strategy: context-window filter kept ${filteredByContext.length}/${eligibleTargets.length} candidates (est. ${estimatedInputTokens} tokens)`
|
|
65
|
+
);
|
|
66
|
+
eligibleTargets = filteredByContext;
|
|
67
|
+
} else {
|
|
68
|
+
log.warn(
|
|
69
|
+
"COMBO",
|
|
70
|
+
`Auto strategy: all candidates filtered by context-window policy (est. ${estimatedInputTokens} tokens), falling back to full pool`
|
|
71
|
+
);
|
|
72
|
+
}
|
|
73
|
+
eligibleTargets = await expandAutoComboCandidatePool(eligibleTargets, combo);
|
|
74
|
+
}
|
|
75
|
+
const prompt = extractPromptForIntent(body);
|
|
76
|
+
const systemPrompt = typeof combo?.system_message === "string" ? combo.system_message : void 0;
|
|
77
|
+
const intentConfig = getIntentConfig(settings, combo);
|
|
78
|
+
const intent = classifyWithConfig(prompt, intentConfig, systemPrompt);
|
|
79
|
+
recordComboIntent(combo.name, intent);
|
|
80
|
+
const taskType = mapIntentToTaskType(intent);
|
|
81
|
+
const {
|
|
82
|
+
routingStrategy,
|
|
83
|
+
candidatePool,
|
|
84
|
+
weights,
|
|
85
|
+
explorationRate,
|
|
86
|
+
budgetCap: configBudgetCap,
|
|
87
|
+
modePack: configModePack,
|
|
88
|
+
resetWindowConfig,
|
|
89
|
+
slaPolicy
|
|
90
|
+
} = parseAutoConfig(combo, eligibleTargets);
|
|
91
|
+
const requestBudgetCap = parseRequestBudgetCap(relayOptions?.budgetCap);
|
|
92
|
+
const budgetCap = requestBudgetCap ?? configBudgetCap;
|
|
93
|
+
const requestModePack = resolveRequestModePack(relayOptions?.mode);
|
|
94
|
+
const modePack = requestModePack.override ? requestModePack.modePack : configModePack;
|
|
95
|
+
if (requestModePack.override || requestBudgetCap !== void 0) {
|
|
96
|
+
log.debug?.(
|
|
97
|
+
"COMBO",
|
|
98
|
+
`Auto strategy: per-request controls applied (mode=${requestModePack.override ? requestModePack.modePack ?? "balanced" : "\u2014"}, budgetCap=${requestBudgetCap ?? "\u2014"})`
|
|
99
|
+
);
|
|
100
|
+
}
|
|
101
|
+
let lastKnownGoodProvider;
|
|
102
|
+
try {
|
|
103
|
+
const { getLKGP } = await import('../../../dist/lib/localDb.js');
|
|
104
|
+
const lkgp = await getLKGP(combo.name, combo.id || combo.name);
|
|
105
|
+
if (lkgp) lastKnownGoodProvider = lkgp.provider;
|
|
106
|
+
} catch (err) {
|
|
107
|
+
log.warn("COMBO", "Failed to retrieve Last Known Good Provider. This is non-fatal.", { err });
|
|
108
|
+
}
|
|
109
|
+
const autoCandidateResilienceSettings = relayOptions?.bypassProviderQuotaPolicy === true ? {
|
|
110
|
+
...resilienceSettings,
|
|
111
|
+
quotaPreflight: {
|
|
112
|
+
...resilienceSettings.quotaPreflight,
|
|
113
|
+
enabled: false
|
|
114
|
+
}
|
|
115
|
+
} : resilienceSettings;
|
|
116
|
+
const candidates = await buildAutoCandidates(
|
|
117
|
+
eligibleTargets,
|
|
118
|
+
combo.name,
|
|
119
|
+
relayOptions?.sessionId,
|
|
120
|
+
resetWindowConfig,
|
|
121
|
+
autoCandidateResilienceSettings
|
|
122
|
+
);
|
|
123
|
+
const routableCandidates = candidates.filter(
|
|
124
|
+
(candidate) => candidate.quotaCutoffBlocked !== true
|
|
125
|
+
);
|
|
126
|
+
const quotaBlockedCount = candidates.length - routableCandidates.length;
|
|
127
|
+
if (quotaBlockedCount > 0) {
|
|
128
|
+
log.info(
|
|
129
|
+
"COMBO",
|
|
130
|
+
`Auto strategy: quota cutoff skipped ${quotaBlockedCount}/${candidates.length} account candidates`
|
|
131
|
+
);
|
|
132
|
+
}
|
|
133
|
+
_registerExecutionCandidates(routableCandidates);
|
|
134
|
+
if (candidates.length > 0 && routableCandidates.length === 0) {
|
|
135
|
+
return {
|
|
136
|
+
earlyResponse: unavailableResponse(
|
|
137
|
+
429,
|
|
138
|
+
"All auto strategy candidates are below configured quota cutoffs"
|
|
139
|
+
)
|
|
140
|
+
};
|
|
141
|
+
}
|
|
142
|
+
if (routableCandidates.length > 0) {
|
|
143
|
+
let selectedProvider = null;
|
|
144
|
+
let selectedModel = null;
|
|
145
|
+
let selectionReason = "";
|
|
146
|
+
if (routingStrategy !== "rules") {
|
|
147
|
+
try {
|
|
148
|
+
const decision = selectWithStrategy(
|
|
149
|
+
routableCandidates,
|
|
150
|
+
{
|
|
151
|
+
taskType,
|
|
152
|
+
requestHasTools,
|
|
153
|
+
lastKnownGoodProvider,
|
|
154
|
+
estimatedInputTokens,
|
|
155
|
+
sla: slaPolicy
|
|
156
|
+
},
|
|
157
|
+
routingStrategy
|
|
158
|
+
);
|
|
159
|
+
selectedProvider = decision.provider;
|
|
160
|
+
selectedModel = decision.model;
|
|
161
|
+
selectionReason = decision.reason;
|
|
162
|
+
autoUsedExplicitRouter = true;
|
|
163
|
+
} catch (err) {
|
|
164
|
+
log.warn(
|
|
165
|
+
"COMBO",
|
|
166
|
+
`Auto strategy '${routingStrategy}' failed (${err?.message || "unknown"}), falling back to rules`
|
|
167
|
+
);
|
|
168
|
+
}
|
|
169
|
+
}
|
|
170
|
+
if (!selectedProvider || !selectedModel) {
|
|
171
|
+
const selection = selectAutoProvider(
|
|
172
|
+
{
|
|
173
|
+
id: combo.id || combo.name,
|
|
174
|
+
name: combo.name,
|
|
175
|
+
type: "auto",
|
|
176
|
+
candidatePool,
|
|
177
|
+
weights,
|
|
178
|
+
modePack,
|
|
179
|
+
budgetCap,
|
|
180
|
+
explorationRate
|
|
181
|
+
},
|
|
182
|
+
routableCandidates,
|
|
183
|
+
taskType
|
|
184
|
+
);
|
|
185
|
+
selectedProvider = selection.provider;
|
|
186
|
+
selectedModel = selection.model;
|
|
187
|
+
selectionReason = `score=${selection.score.toFixed(3)}${selection.isExploration ? " (exploration)" : ""}`;
|
|
188
|
+
}
|
|
189
|
+
const autoManifestHint = config.complexityAwareRouting === true ? buildComplexityRoutingHint(
|
|
190
|
+
eligibleTargets.filter((t) => t.kind === "model"),
|
|
191
|
+
body,
|
|
192
|
+
log
|
|
193
|
+
) : null;
|
|
194
|
+
const scoredTargets = scoreAutoTargets(
|
|
195
|
+
eligibleTargets,
|
|
196
|
+
routableCandidates,
|
|
197
|
+
taskType,
|
|
198
|
+
weights,
|
|
199
|
+
autoManifestHint
|
|
200
|
+
);
|
|
201
|
+
const rankedTargets = scoredTargets.map((entry) => entry.target);
|
|
202
|
+
const selectedTarget = scoredTargets.find((entry) => {
|
|
203
|
+
const parsed = parseModel(entry.target.modelStr);
|
|
204
|
+
const modelId = parsed.model || entry.target.modelStr;
|
|
205
|
+
return entry.target.provider === selectedProvider && modelId === selectedModel;
|
|
206
|
+
})?.target || rankedTargets[0] || eligibleTargets[0];
|
|
207
|
+
if (!selectedTarget) {
|
|
208
|
+
return {
|
|
209
|
+
earlyResponse: unavailableResponse(
|
|
210
|
+
429,
|
|
211
|
+
"No auto strategy targets remained after quota cutoff filtering"
|
|
212
|
+
)
|
|
213
|
+
};
|
|
214
|
+
}
|
|
215
|
+
orderedTargets = dedupeTargetsByExecutionKey(
|
|
216
|
+
[selectedTarget, ...rankedTargets, ...eligibleTargets].filter(
|
|
217
|
+
(entry) => entry !== void 0 && entry !== null
|
|
218
|
+
)
|
|
219
|
+
);
|
|
220
|
+
log.info(
|
|
221
|
+
"COMBO",
|
|
222
|
+
`Auto selection: ${selectedTarget?.modelStr || `${selectedProvider}/${selectedModel}`} | intent=${intent} task=${taskType} | strategy=${routingStrategy} | ${selectionReason}`
|
|
223
|
+
);
|
|
224
|
+
} else {
|
|
225
|
+
log.warn("COMBO", "Auto strategy has no candidates, keeping default ordering");
|
|
226
|
+
}
|
|
227
|
+
return { orderedTargets, autoUsedExplicitRouter };
|
|
228
|
+
}
|
|
229
|
+
export {
|
|
230
|
+
resolveAutoStrategyOrder
|
|
231
|
+
};
|