adaptive-memory-multi-model-router 2.16.0 → 2.16.1
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/.github/workflows/adapters-ci.yml +142 -0
- package/.github/workflows/auto-submit-sitemap.yml +41 -0
- package/.github/workflows/ci.yml +2 -5
- package/.github/workflows/mcp-pypi-publish.yml +34 -0
- package/.github/workflows/pypi-publish.yml +146 -0
- package/.github/workflows/tmlpd-publish.yml +23 -0
- package/README.md +245 -148
- package/RELEASE_v2.16.0.md +149 -0
- package/TECHNICAL_README.md +253 -0
- package/adapters/README.md +36 -0
- package/adapters/__init__.py +25 -0
- package/adapters/a3m_adapter/__init__.py +51 -0
- package/adapters/a3m_adapter/adapter/__init__.py +22 -0
- package/adapters/a3m_adapter/adapter/autogen.py +169 -0
- package/adapters/a3m_adapter/adapter/config.py +100 -0
- package/adapters/a3m_adapter/adapter/haystack.py +197 -0
- package/adapters/a3m_adapter/adapter/langchain.py +155 -0
- package/adapters/a3m_adapter/adapter/langgraph.py +196 -0
- package/adapters/a3m_adapter/adapter/llamaindex.py +162 -0
- package/adapters/a3m_adapter/adapter/pinecone.py +217 -0
- package/adapters/a3m_adapter/adapter/vercel.py +188 -0
- package/adapters/a3m_adapter/tests/__init__.py +1 -0
- package/adapters/a3m_adapter/tests/test_adapters.py +118 -0
- package/adapters/a3m_adapter/tests/test_integration.py +80 -0
- package/adapters/requirements-dev.txt +6 -0
- package/adapters/requirements.txt +4 -0
- package/adapters/setup.py +23 -0
- package/demo.py +251 -0
- package/discoverability-diagnosis.md +280 -0
- package/dist/analytics/costAnalytics.d.ts.map +1 -1
- package/dist/benchmark/reproducible.d.ts.map +1 -1
- package/dist/cache/semanticCache.d.ts.map +1 -1
- package/dist/cli/setupWizard.d.ts +257 -50
- package/dist/cli/setupWizard.d.ts.map +1 -1
- package/dist/cli/setupWizard.js +419 -109
- package/dist/cli/setupWizard.js.map +1 -1
- package/dist/cli/tui.js +96 -67
- package/dist/cli.js +9 -0
- package/dist/cost/budgetEnforcer.d.ts.map +1 -1
- package/dist/cost/costTracker.d.ts.map +1 -1
- package/dist/ensemble/multiRoundDialog.d.ts.map +1 -1
- package/dist/ensemble/shapleyValue.d.ts.map +1 -1
- package/dist/ensemble.d.ts +1 -1
- package/dist/ensemble.js +141 -0
- package/dist/integrations/langchainAdapter.d.ts.map +1 -1
- package/dist/integrations/langchainAdapter.js +3 -3
- package/dist/integrations/langchainAdapter.js.map +1 -1
- package/dist/integrations/oauth.d.ts.map +1 -1
- package/dist/integrations/scienceAdapter.d.ts.map +1 -1
- package/dist/memory/autoFetch.d.ts.map +1 -1
- package/dist/memory/hybridMemory.d.ts.map +1 -1
- package/dist/memory/memoryTree.d.ts.map +1 -1
- package/dist/memory/obsidianVault.d.ts.map +1 -1
- package/dist/memory/reasoningBank.d.ts.map +1 -1
- package/dist/observability/metrics.d.ts.map +1 -1
- package/dist/observability/tracer.d.ts.map +1 -1
- package/dist/providers/providerConfig.d.ts.map +1 -1
- package/dist/providers/providerConfig.js +32 -17
- package/dist/providers/providerConfig.js.map +1 -1
- package/dist/routing/advancedRouter.d.ts.map +1 -1
- package/dist/routing/advancedRouter.js +106 -14
- package/dist/routing/advancedRouter.js.map +1 -1
- package/dist/routing/providerHealth.d.ts.map +1 -1
- package/dist/routing/providerRetry.d.ts.map +1 -1
- package/dist/routing/shadowSampler.js.map +1 -1
- package/dist/security/guardrails.d.ts.map +1 -1
- package/dist/server/handlers/chatHandler.d.ts.map +1 -1
- package/dist/server/handlers/completionsHandler.d.ts.map +1 -1
- package/dist/server/handlers/embeddingsHandler.d.ts.map +1 -1
- package/dist/server/handlers/healthHandler.d.ts.map +1 -1
- package/dist/server/handlers/metricsHandler.d.ts.map +1 -1
- package/dist/server/handlers/modelsHandler.d.ts.map +1 -1
- package/dist/server/metrics.d.ts.map +1 -1
- package/dist/server/proxyServer.d.ts.map +1 -1
- package/dist/server/router.d.ts.map +1 -1
- package/dist/server/state.d.ts.map +1 -1
- package/dist/skills/__tests__/skill_manager.test.js +5 -265
- package/dist/skills/__tests__/skill_manager.test.js.map +1 -1
- package/dist/utils/tokenUtils.d.ts.map +1 -1
- package/docker-compose.yml +84 -60
- package/docs/ARTICLE_Biology_Inspired_Routing.md +208 -0
- package/docs/ARTICLE_Master.md +78 -0
- package/docs/ARTICLE_Master_CN.md +78 -0
- package/docs/ARTICLE_OpenRouter_Stripe.md +140 -0
- package/docs/DEVPTO_ARTICLE.md +84 -0
- package/docs/HUMAN_STYLE_GUIDE.md +75 -0
- package/docs/IMPRINT_PLAN.md +88 -0
- package/docs/OPENROUTER_ALTERNATIVE.md +184 -0
- package/docs/SOCIAL_CAMPAIGN.md +316 -0
- package/docs/anthropic.html +45 -0
- package/docs/best-llm-routers-2025.html +157 -0
- package/docs/cerebras.html +43 -0
- package/docs/cli-cheatsheet.md +286 -212
- package/docs/comparison.md +2 -2
- package/docs/deepseek.html +44 -0
- package/docs/google.html +47 -0
- package/docs/groq.html +44 -0
- package/docs/llms-full.txt +360 -138
- package/docs/llms.txt +70 -71
- package/docs/mistral.html +43 -0
- package/docs/ollama.html +50 -0
- package/docs/openai.html +57 -0
- package/docs/sitemap.xml +69 -57
- package/docs-site/blog/best-llm-routers-2025.html +157 -0
- package/docs-site/index.html +68 -9
- package/docs-site/providers/anthropic.html +45 -0
- package/docs-site/providers/cerebras.html +43 -0
- package/docs-site/providers/deepseek.html +44 -0
- package/docs-site/providers/google.html +47 -0
- package/docs-site/providers/groq.html +44 -0
- package/docs-site/providers/index.html +41 -0
- package/docs-site/providers/mistral.html +43 -0
- package/docs-site/providers/ollama.html +50 -0
- package/docs-site/providers/openai.html +57 -0
- package/docs-site/sitemap.xml +69 -0
- package/llms.txt +70 -62
- package/package.json +44 -182
- package/packages/agentkit-adapter/LICENSE +21 -0
- package/packages/agentkit-adapter/README.md +126 -0
- package/packages/agentkit-adapter/examples/agentkit-example.ts +139 -0
- package/packages/agentkit-adapter/package.json +57 -0
- package/packages/agentkit-adapter/src/adapter.ts +381 -0
- package/packages/agentkit-adapter/src/index.ts +36 -0
- package/packages/agentkit-adapter/src/types.ts +105 -0
- package/packages/agentkit-adapter/src/util.ts +13 -0
- package/packages/agentkit-adapter/tsconfig.json +22 -0
- package/prometheus.yml +8 -0
- package/python/README.md +35 -81
- package/python/a3m/__init__.py +32 -3
- package/python/a3m/adapters/__init__.py +21 -0
- package/python/a3m/adapters/langchain.py +190 -0
- package/python/a3m/adapters/llamaindex.py +249 -0
- package/python/a3m/adapters/qdrant.py +240 -0
- package/python/a3m/adapters/weaviate.py +263 -0
- package/python/a3m/client.py +5 -0
- package/python/build_verify.sh +32 -0
- package/python/mcp-server/README.md +172 -0
- package/python/mcp-server/a3m_mcp/__init__.py +25 -0
- package/python/mcp-server/a3m_mcp/__main__.py +15 -0
- package/python/mcp-server/a3m_mcp/server.py +205 -0
- package/python/mcp-server/pyproject.toml +24 -0
- package/python/pyproject.toml +60 -6
- package/python/setup.py +3 -28
- package/scripts/submit-sitemap.sh +52 -0
- package/src/__types__/registry.d.ts +14 -0
- package/src/cli/setupWizard.ts +443 -112
- package/src/cli/tui.ts +159 -0
- package/src/ensemble.ts +154 -1
- package/src/integrations/langchainAdapter.ts +2 -2
- package/src/providers/providerConfig.ts +32 -17
- package/src/providers/registry.js +27 -0
- package/src/routing/advancedRouter.ts +99 -14
- package/src/routing/shadowSampler.ts +1 -1
- package/test-install/package.json +12 -0
- package/tests/tsconfig.json +0 -1
- package/tmlpd-pi-extension/README.md +105 -44
- package/tmlpd-pi-extension/docs/demo.svg +33 -0
- package/tmlpd-pi-extension/package.json +35 -106
- package/tmlpd-pi-extension/src/tokenOptimization/contextStratifier.ts +163 -0
- package/tmlpd-pi-extension/src/tokenOptimization/fetchOnceLocal.ts +136 -0
- package/tmlpd-pi-extension/src/tokenOptimization/index.ts +197 -0
- package/tmlpd-pi-extension/src/tokenOptimization/interAgentCompression.ts +157 -0
- package/tmlpd-pi-extension/src/tokenOptimization/schemaContract.ts +101 -0
- package/tmlpd-pi-extension/src/tokenOptimization/semanticCache.ts +248 -0
- package/tmlpd-pi-extension/src/tokenOptimization/tokenAwareFallback.ts +192 -0
- package/tmlpd-pi-extension/test/verify.js +21 -0
- package/tsconfig.build.json +3 -2
- package/packages/a3m-vercel-ai/dist/a3m-language-model.d.ts +0 -12
- package/packages/a3m-vercel-ai/dist/a3m-language-model.d.ts.map +0 -1
- package/packages/a3m-vercel-ai/dist/a3m-language-model.js +0 -289
- package/packages/a3m-vercel-ai/dist/a3m-language-model.js.map +0 -1
- package/packages/a3m-vercel-ai/dist/index.d.ts +0 -82
- package/packages/a3m-vercel-ai/dist/index.d.ts.map +0 -1
- package/packages/a3m-vercel-ai/dist/index.js +0 -79
- package/packages/a3m-vercel-ai/dist/index.js.map +0 -1
- package/packages/a3m-vercel-ai/dist/types.d.ts +0 -97
- package/packages/a3m-vercel-ai/dist/types.d.ts.map +0 -1
- package/packages/a3m-vercel-ai/dist/types.js +0 -5
- package/packages/a3m-vercel-ai/dist/types.js.map +0 -1
- package/src/skills/__tests__/skill_manager.test.ts +0 -328
- package/tmlpd-pi-extension/dist/cache/prefixCache.d.ts +0 -114
- package/tmlpd-pi-extension/dist/cache/prefixCache.d.ts.map +0 -1
- package/tmlpd-pi-extension/dist/cache/prefixCache.js +0 -285
- package/tmlpd-pi-extension/dist/cache/prefixCache.js.map +0 -1
- package/tmlpd-pi-extension/dist/cache/responseCache.d.ts +0 -58
- package/tmlpd-pi-extension/dist/cache/responseCache.d.ts.map +0 -1
- package/tmlpd-pi-extension/dist/cache/responseCache.js +0 -153
- package/tmlpd-pi-extension/dist/cache/responseCache.js.map +0 -1
- package/tmlpd-pi-extension/dist/cli.js +0 -59
- package/tmlpd-pi-extension/dist/cost/costTracker.d.ts +0 -95
- package/tmlpd-pi-extension/dist/cost/costTracker.d.ts.map +0 -1
- package/tmlpd-pi-extension/dist/cost/costTracker.js +0 -240
- package/tmlpd-pi-extension/dist/cost/costTracker.js.map +0 -1
- package/tmlpd-pi-extension/dist/index.d.ts +0 -723
- package/tmlpd-pi-extension/dist/index.d.ts.map +0 -1
- package/tmlpd-pi-extension/dist/index.js +0 -239
- package/tmlpd-pi-extension/dist/index.js.map +0 -1
- package/tmlpd-pi-extension/dist/memory/episodicMemory.d.ts +0 -82
- package/tmlpd-pi-extension/dist/memory/episodicMemory.d.ts.map +0 -1
- package/tmlpd-pi-extension/dist/memory/episodicMemory.js +0 -145
- package/tmlpd-pi-extension/dist/memory/episodicMemory.js.map +0 -1
- package/tmlpd-pi-extension/dist/orchestration/haloOrchestrator.d.ts +0 -102
- package/tmlpd-pi-extension/dist/orchestration/haloOrchestrator.d.ts.map +0 -1
- package/tmlpd-pi-extension/dist/orchestration/haloOrchestrator.js +0 -207
- package/tmlpd-pi-extension/dist/orchestration/haloOrchestrator.js.map +0 -1
- package/tmlpd-pi-extension/dist/orchestration/mctsWorkflow.d.ts +0 -85
- package/tmlpd-pi-extension/dist/orchestration/mctsWorkflow.d.ts.map +0 -1
- package/tmlpd-pi-extension/dist/orchestration/mctsWorkflow.js +0 -210
- package/tmlpd-pi-extension/dist/orchestration/mctsWorkflow.js.map +0 -1
- package/tmlpd-pi-extension/dist/providers/localProvider.d.ts +0 -102
- package/tmlpd-pi-extension/dist/providers/localProvider.d.ts.map +0 -1
- package/tmlpd-pi-extension/dist/providers/localProvider.js +0 -338
- package/tmlpd-pi-extension/dist/providers/localProvider.js.map +0 -1
- package/tmlpd-pi-extension/dist/providers/registry.d.ts +0 -55
- package/tmlpd-pi-extension/dist/providers/registry.d.ts.map +0 -1
- package/tmlpd-pi-extension/dist/providers/registry.js +0 -138
- package/tmlpd-pi-extension/dist/providers/registry.js.map +0 -1
- package/tmlpd-pi-extension/dist/routing/advancedRouter.d.ts +0 -68
- package/tmlpd-pi-extension/dist/routing/advancedRouter.d.ts.map +0 -1
- package/tmlpd-pi-extension/dist/routing/advancedRouter.js +0 -332
- package/tmlpd-pi-extension/dist/routing/advancedRouter.js.map +0 -1
- package/tmlpd-pi-extension/dist/tools/tmlpdTools.d.ts +0 -101
- package/tmlpd-pi-extension/dist/tools/tmlpdTools.d.ts.map +0 -1
- package/tmlpd-pi-extension/dist/tools/tmlpdTools.js +0 -368
- package/tmlpd-pi-extension/dist/tools/tmlpdTools.js.map +0 -1
- package/tmlpd-pi-extension/dist/utils/batchProcessor.d.ts +0 -96
- package/tmlpd-pi-extension/dist/utils/batchProcessor.d.ts.map +0 -1
- package/tmlpd-pi-extension/dist/utils/batchProcessor.js +0 -170
- package/tmlpd-pi-extension/dist/utils/batchProcessor.js.map +0 -1
- package/tmlpd-pi-extension/dist/utils/compression.d.ts +0 -61
- package/tmlpd-pi-extension/dist/utils/compression.d.ts.map +0 -1
- package/tmlpd-pi-extension/dist/utils/compression.js +0 -281
- package/tmlpd-pi-extension/dist/utils/compression.js.map +0 -1
- package/tmlpd-pi-extension/dist/utils/reliability.d.ts +0 -74
- package/tmlpd-pi-extension/dist/utils/reliability.d.ts.map +0 -1
- package/tmlpd-pi-extension/dist/utils/reliability.js +0 -177
- package/tmlpd-pi-extension/dist/utils/reliability.js.map +0 -1
- package/tmlpd-pi-extension/dist/utils/speculativeDecoding.d.ts +0 -117
- package/tmlpd-pi-extension/dist/utils/speculativeDecoding.d.ts.map +0 -1
- package/tmlpd-pi-extension/dist/utils/speculativeDecoding.js +0 -246
- package/tmlpd-pi-extension/dist/utils/speculativeDecoding.js.map +0 -1
- package/tmlpd-pi-extension/dist/utils/tokenUtils.d.ts +0 -50
- package/tmlpd-pi-extension/dist/utils/tokenUtils.d.ts.map +0 -1
- package/tmlpd-pi-extension/dist/utils/tokenUtils.js +0 -124
- package/tmlpd-pi-extension/dist/utils/tokenUtils.js.map +0 -1
|
@@ -0,0 +1,192 @@
|
|
|
1
|
+
/**
|
|
2
|
+
* Token-Aware Fallback Chain - Token Optimization Pattern #3
|
|
3
|
+
*
|
|
4
|
+
* Route to smaller models when high-token content is detected.
|
|
5
|
+
* Reduces cost by using cheap models for simple/high-volume queries.
|
|
6
|
+
*
|
|
7
|
+
* Based on: arXiv:2608.17188 - Token Optimization and Context Window Management
|
|
8
|
+
*/
|
|
9
|
+
|
|
10
|
+
export interface TokenThreshold {
|
|
11
|
+
maxTokens: number;
|
|
12
|
+
models: string[]; // Available models at this tier
|
|
13
|
+
}
|
|
14
|
+
|
|
15
|
+
export interface TokenAwareFallbackConfig {
|
|
16
|
+
thresholds: TokenThreshold[];
|
|
17
|
+
defaultModel?: string;
|
|
18
|
+
enableDynamicThreshold?: boolean;
|
|
19
|
+
}
|
|
20
|
+
|
|
21
|
+
export interface FallbackDecision {
|
|
22
|
+
targetModel: string;
|
|
23
|
+
tier: "cheap" | "medium" | "expensive";
|
|
24
|
+
estimatedTokens: number;
|
|
25
|
+
reasoning: string;
|
|
26
|
+
}
|
|
27
|
+
|
|
28
|
+
/**
|
|
29
|
+
* Model cost tiers (approximate, per 1M tokens)
|
|
30
|
+
*/
|
|
31
|
+
const MODEL_COSTS: Record<string, number> = {
|
|
32
|
+
// Cheap tier (< $0.50/1M)
|
|
33
|
+
"gpt-4o-mini": 0.15,
|
|
34
|
+
"claude-haiku": 0.25,
|
|
35
|
+
"gemini-flash": 0.075,
|
|
36
|
+
"gpt-3.5-turbo": 0.50,
|
|
37
|
+
|
|
38
|
+
// Medium tier ($0.50 - $3.00/1M)
|
|
39
|
+
"gpt-4o": 2.50,
|
|
40
|
+
"claude-sonnet": 3.00,
|
|
41
|
+
"gemini-pro": 1.25,
|
|
42
|
+
|
|
43
|
+
// Expensive tier (> $3.00/1M)
|
|
44
|
+
"gpt-4-turbo": 10.00,
|
|
45
|
+
"claude-opus": 15.00,
|
|
46
|
+
"gpt-4": 30.00,
|
|
47
|
+
};
|
|
48
|
+
|
|
49
|
+
export class TokenAwareFallback {
|
|
50
|
+
private config: TokenAwareFallbackConfig;
|
|
51
|
+
|
|
52
|
+
constructor(config: Partial<TokenAwareFallbackConfig> = {}) {
|
|
53
|
+
this.config = {
|
|
54
|
+
thresholds: config.thresholds ?? [
|
|
55
|
+
{ maxTokens: 500, models: ["gpt-4o-mini", "claude-haiku", "gemini-flash"] },
|
|
56
|
+
{ maxTokens: 2000, models: ["gpt-4o", "claude-sonnet", "gemini-pro"] },
|
|
57
|
+
{ maxTokens: Infinity, models: ["gpt-4o", "claude-opus", "gpt-4-turbo"] },
|
|
58
|
+
],
|
|
59
|
+
defaultModel: config.defaultModel ?? "gpt-4o",
|
|
60
|
+
enableDynamicThreshold: config.enableDynamicThreshold ?? false,
|
|
61
|
+
};
|
|
62
|
+
}
|
|
63
|
+
|
|
64
|
+
/**
|
|
65
|
+
* Estimate token count from text (rough approximation)
|
|
66
|
+
*/
|
|
67
|
+
estimateTokens(text: string): number {
|
|
68
|
+
// ~4 characters per token for English
|
|
69
|
+
return Math.ceil(text.length / 4);
|
|
70
|
+
}
|
|
71
|
+
|
|
72
|
+
/**
|
|
73
|
+
* Estimate tokens including history
|
|
74
|
+
*/
|
|
75
|
+
estimateTotalTokens(
|
|
76
|
+
query: string,
|
|
77
|
+
history?: Array<{ content: string }>
|
|
78
|
+
): number {
|
|
79
|
+
let total = this.estimateTokens(query);
|
|
80
|
+
|
|
81
|
+
if (history) {
|
|
82
|
+
total += history.reduce((sum, msg) => sum + this.estimateTokens(msg.content), 0);
|
|
83
|
+
}
|
|
84
|
+
|
|
85
|
+
return total;
|
|
86
|
+
}
|
|
87
|
+
|
|
88
|
+
/**
|
|
89
|
+
* Select appropriate model based on token count
|
|
90
|
+
*/
|
|
91
|
+
selectModel(estimatedTokens: number): FallbackDecision {
|
|
92
|
+
for (const threshold of this.config.thresholds) {
|
|
93
|
+
if (estimatedTokens <= threshold.maxTokens) {
|
|
94
|
+
const model = threshold.models[0]; // Select first available in tier
|
|
95
|
+
const tier = this.getModelTier(model);
|
|
96
|
+
|
|
97
|
+
return {
|
|
98
|
+
targetModel: model,
|
|
99
|
+
tier,
|
|
100
|
+
estimatedTokens,
|
|
101
|
+
reasoning: `Selected ${tier} model ${model} for ${estimatedTokens} tokens`,
|
|
102
|
+
};
|
|
103
|
+
}
|
|
104
|
+
}
|
|
105
|
+
|
|
106
|
+
// Fallback to default
|
|
107
|
+
return {
|
|
108
|
+
targetModel: this.config.defaultModel!,
|
|
109
|
+
tier: "medium",
|
|
110
|
+
estimatedTokens,
|
|
111
|
+
reasoning: `Defaulting to ${this.config.defaultModel} for ${estimatedTokens} tokens`,
|
|
112
|
+
};
|
|
113
|
+
}
|
|
114
|
+
|
|
115
|
+
/**
|
|
116
|
+
* Get cost tier for a model
|
|
117
|
+
*/
|
|
118
|
+
private getModelTier(model: string): "cheap" | "medium" | "expensive" {
|
|
119
|
+
const cost = MODEL_COSTS[model.toLowerCase()] ?? 1.0;
|
|
120
|
+
if (cost < 0.5) return "cheap";
|
|
121
|
+
if (cost < 3.0) return "medium";
|
|
122
|
+
return "expensive";
|
|
123
|
+
}
|
|
124
|
+
|
|
125
|
+
/**
|
|
126
|
+
* Create fallback chain for a query
|
|
127
|
+
*/
|
|
128
|
+
createFallbackChain(
|
|
129
|
+
query: string,
|
|
130
|
+
history?: Array<{ content: string }>
|
|
131
|
+
): string[] {
|
|
132
|
+
const estimatedTokens = this.estimateTotalTokens(query, history);
|
|
133
|
+
const chain: string[] = [];
|
|
134
|
+
|
|
135
|
+
// Find appropriate tier and all models in that tier
|
|
136
|
+
for (const threshold of this.config.thresholds) {
|
|
137
|
+
if (estimatedTokens <= threshold.maxTokens) {
|
|
138
|
+
chain.push(...threshold.models);
|
|
139
|
+
break;
|
|
140
|
+
}
|
|
141
|
+
}
|
|
142
|
+
|
|
143
|
+
// If still empty, add all models
|
|
144
|
+
if (chain.length === 0) {
|
|
145
|
+
for (const threshold of this.config.thresholds) {
|
|
146
|
+
chain.push(...threshold.models);
|
|
147
|
+
}
|
|
148
|
+
}
|
|
149
|
+
|
|
150
|
+
return chain;
|
|
151
|
+
}
|
|
152
|
+
|
|
153
|
+
/**
|
|
154
|
+
* Estimate cost savings vs always using expensive model
|
|
155
|
+
*/
|
|
156
|
+
estimateSavings(
|
|
157
|
+
query: string,
|
|
158
|
+
history?: Array<{ content: string }>
|
|
159
|
+
): { savingsPercent: number; cheapModel: string; expensiveModel: string } {
|
|
160
|
+
const estimatedTokens = this.estimateTotalTokens(query, history);
|
|
161
|
+
const decision = this.selectModel(estimatedTokens);
|
|
162
|
+
|
|
163
|
+
const cheapCost = this.getModelCost(decision.targetModel) * estimatedTokens / 1_000_000;
|
|
164
|
+
const expensiveCost = this.getModelCost("gpt-4-turbo") * estimatedTokens / 1_000_000;
|
|
165
|
+
|
|
166
|
+
const savingsPercent = expensiveCost > 0
|
|
167
|
+
? ((expensiveCost - cheapCost) / expensiveCost) * 100
|
|
168
|
+
: 0;
|
|
169
|
+
|
|
170
|
+
return {
|
|
171
|
+
savingsPercent: Math.round(savingsPercent * 100) / 100,
|
|
172
|
+
cheapModel: decision.targetModel,
|
|
173
|
+
expensiveModel: "gpt-4-turbo",
|
|
174
|
+
};
|
|
175
|
+
}
|
|
176
|
+
|
|
177
|
+
/**
|
|
178
|
+
* Get model cost per 1M tokens
|
|
179
|
+
*/
|
|
180
|
+
getModelCost(model: string): number {
|
|
181
|
+
return MODEL_COSTS[model.toLowerCase()] ?? 1.0;
|
|
182
|
+
}
|
|
183
|
+
|
|
184
|
+
/**
|
|
185
|
+
* Get all configured thresholds
|
|
186
|
+
*/
|
|
187
|
+
getThresholds(): TokenThreshold[] {
|
|
188
|
+
return this.config.thresholds;
|
|
189
|
+
}
|
|
190
|
+
}
|
|
191
|
+
|
|
192
|
+
export default TokenAwareFallback;
|
|
@@ -0,0 +1,21 @@
|
|
|
1
|
+
// Smoke test: package compiles, core exports + tokenOptimization module load
|
|
2
|
+
const assert = require("assert");
|
|
3
|
+
|
|
4
|
+
const mod = require("../dist/index.js");
|
|
5
|
+
|
|
6
|
+
// Core exports exist
|
|
7
|
+
for (const name of ["createTMLPD", "TMLPDTools", "ResponseCache", "CostTracker", "ProviderRegistry"]) {
|
|
8
|
+
assert(name in mod, `missing core export: ${name}`);
|
|
9
|
+
}
|
|
10
|
+
|
|
11
|
+
// tokenOptimization sub-module loads (added v1.3.2)
|
|
12
|
+
const to = require("../dist/tokenOptimization/index.js");
|
|
13
|
+
for (const name of ["SemanticCache", "ContextStratifier", "TokenAwareFallback", "SchemaContractor", "FetchOnceProcessor"]) {
|
|
14
|
+
assert(name in to, `missing tokenOptimization export: ${name}`);
|
|
15
|
+
}
|
|
16
|
+
|
|
17
|
+
// SemanticCache basic behavior
|
|
18
|
+
const cache = new to.SemanticCache({ maxEntries: 10 });
|
|
19
|
+
assert(typeof cache.get === "function" && typeof cache.set === "function", "SemanticCache API broken");
|
|
20
|
+
|
|
21
|
+
console.log("✓ smoke test passed — dist loads, tokenOptimization exports verified");
|
package/tsconfig.build.json
CHANGED
|
@@ -13,8 +13,7 @@
|
|
|
13
13
|
"sourceMap": true,
|
|
14
14
|
"types": ["node"],
|
|
15
15
|
"resolveJsonModule": true,
|
|
16
|
-
"moduleResolution": "node"
|
|
17
|
-
"ignoreDeprecations": "6.0"
|
|
16
|
+
"moduleResolution": "node"
|
|
18
17
|
},
|
|
19
18
|
"include": ["src/**/*"],
|
|
20
19
|
"exclude": [
|
|
@@ -22,7 +21,9 @@
|
|
|
22
21
|
"dist",
|
|
23
22
|
"src/skills/__tests__",
|
|
24
23
|
"src/cli/setupWizard.ts",
|
|
24
|
+
"src/cli/tui.ts",
|
|
25
25
|
"src/integrations/langchainAdapter.ts",
|
|
26
|
+
"src/tui/dashboard.ts",
|
|
26
27
|
"src/tui/index.ts",
|
|
27
28
|
"src/benchmark/comprehensive.ts"
|
|
28
29
|
]
|
|
@@ -1,12 +0,0 @@
|
|
|
1
|
-
/**
|
|
2
|
-
* A3M Router Language Model for Vercel AI SDK (v3)
|
|
3
|
-
*
|
|
4
|
-
* Implements the LanguageModelV1 interface from @ai-sdk/provider
|
|
5
|
-
*/
|
|
6
|
-
import { LanguageModelV1 } from '@ai-sdk/provider';
|
|
7
|
-
import type { A3MRouterConfig } from './types';
|
|
8
|
-
/**
|
|
9
|
-
* Create an A3M Router Language Model for Vercel AI SDK
|
|
10
|
-
*/
|
|
11
|
-
export declare function createA3MLanguageModel(config?: A3MRouterConfig): LanguageModelV1;
|
|
12
|
-
//# sourceMappingURL=a3m-language-model.d.ts.map
|
|
@@ -1 +0,0 @@
|
|
|
1
|
-
{"version":3,"file":"a3m-language-model.d.ts","sourceRoot":"","sources":["../src/a3m-language-model.ts"],"names":[],"mappings":"AAAA;;;;GAIG;AAEH,OAAO,EACL,eAAe,EAOhB,MAAM,kBAAkB,CAAC;AAC1B,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,SAAS,CAAC;AA6C/C;;GAEG;AACH,wBAAgB,sBAAsB,CAAC,MAAM,GAAE,eAAoB,mBA6TlE"}
|
|
@@ -1,289 +0,0 @@
|
|
|
1
|
-
/**
|
|
2
|
-
* A3M Router Language Model for Vercel AI SDK (v3)
|
|
3
|
-
*
|
|
4
|
-
* Implements the LanguageModelV1 interface from @ai-sdk/provider
|
|
5
|
-
*/
|
|
6
|
-
/**
|
|
7
|
-
* Convert Vercel AI SDK prompt format to OpenAI format for A3M
|
|
8
|
-
*/
|
|
9
|
-
function convertMessages(prompt) {
|
|
10
|
-
const result = [];
|
|
11
|
-
// Handle string prompts
|
|
12
|
-
if (typeof prompt === 'string') {
|
|
13
|
-
return [{ role: 'user', content: prompt }];
|
|
14
|
-
}
|
|
15
|
-
// messages is an array of prompt parts
|
|
16
|
-
for (const part of prompt) {
|
|
17
|
-
if (part.role === 'system') {
|
|
18
|
-
result.push({ role: 'system', content: part.content });
|
|
19
|
-
}
|
|
20
|
-
else if (part.role === 'user') {
|
|
21
|
-
if (typeof part.content === 'string') {
|
|
22
|
-
result.push({ role: 'user', content: part.content });
|
|
23
|
-
}
|
|
24
|
-
else if (Array.isArray(part.content)) {
|
|
25
|
-
const textContent = part.content
|
|
26
|
-
.filter((p) => p.type === 'text')
|
|
27
|
-
.map(p => p.text)
|
|
28
|
-
.join('\n');
|
|
29
|
-
result.push({ role: 'user', content: textContent });
|
|
30
|
-
}
|
|
31
|
-
}
|
|
32
|
-
else if (part.role === 'assistant') {
|
|
33
|
-
if (typeof part.content === 'string') {
|
|
34
|
-
result.push({ role: 'assistant', content: part.content });
|
|
35
|
-
}
|
|
36
|
-
else if (Array.isArray(part.content)) {
|
|
37
|
-
const textContent = part.content
|
|
38
|
-
.filter((p) => p.type === 'text')
|
|
39
|
-
.map(p => p.text)
|
|
40
|
-
.join('\n');
|
|
41
|
-
result.push({ role: 'assistant', content: textContent });
|
|
42
|
-
}
|
|
43
|
-
}
|
|
44
|
-
}
|
|
45
|
-
return result;
|
|
46
|
-
}
|
|
47
|
-
/**
|
|
48
|
-
* Create an A3M Router Language Model for Vercel AI SDK
|
|
49
|
-
*/
|
|
50
|
-
export function createA3MLanguageModel(config = {}) {
|
|
51
|
-
const baseURL = config.baseURL || 'http://localhost:8787';
|
|
52
|
-
const apiKey = config.apiKey || 'not-needed';
|
|
53
|
-
const model = {
|
|
54
|
-
specificationVersion: 'v1',
|
|
55
|
-
provider: 'a3m',
|
|
56
|
-
modelId: 'auto',
|
|
57
|
-
defaultObjectGenerationMode: undefined,
|
|
58
|
-
async doGenerate(options) {
|
|
59
|
-
const messages = convertMessages(options.prompt);
|
|
60
|
-
// Extract system prompt if present
|
|
61
|
-
const systemMessages = messages.filter(m => m.role === 'system');
|
|
62
|
-
const nonSystemMessages = messages.filter(m => m.role !== 'system');
|
|
63
|
-
const system = systemMessages.map(m => m.content).join('\n');
|
|
64
|
-
// Build the request body for A3M
|
|
65
|
-
const requestBody = {
|
|
66
|
-
model: 'auto',
|
|
67
|
-
messages: nonSystemMessages.length > 0 ? nonSystemMessages : [{ role: 'user', content: ' ' }],
|
|
68
|
-
};
|
|
69
|
-
if (system) {
|
|
70
|
-
requestBody.system = system;
|
|
71
|
-
}
|
|
72
|
-
if (options.temperature !== undefined) {
|
|
73
|
-
requestBody.temperature = options.temperature;
|
|
74
|
-
}
|
|
75
|
-
if (options.maxTokens !== undefined) {
|
|
76
|
-
requestBody.max_tokens = options.maxTokens;
|
|
77
|
-
}
|
|
78
|
-
if (options.topP !== undefined) {
|
|
79
|
-
requestBody.top_p = options.topP;
|
|
80
|
-
}
|
|
81
|
-
if (options.stopSequences !== undefined) {
|
|
82
|
-
requestBody.stop = options.stopSequences;
|
|
83
|
-
}
|
|
84
|
-
if (config.parallelEnsemble) {
|
|
85
|
-
requestBody.parallel = config.parallelCount || 3;
|
|
86
|
-
}
|
|
87
|
-
if (config.stealth) {
|
|
88
|
-
requestBody.stealth = true;
|
|
89
|
-
}
|
|
90
|
-
try {
|
|
91
|
-
// Call A3M Router
|
|
92
|
-
const response = await fetch(`${baseURL}/v1/chat/completions`, {
|
|
93
|
-
method: 'POST',
|
|
94
|
-
headers: {
|
|
95
|
-
'Content-Type': 'application/json',
|
|
96
|
-
'Authorization': `Bearer ${apiKey}`,
|
|
97
|
-
},
|
|
98
|
-
body: JSON.stringify(requestBody),
|
|
99
|
-
});
|
|
100
|
-
if (!response.ok) {
|
|
101
|
-
const error = await response.text();
|
|
102
|
-
throw new Error(`A3M Router error: ${response.status} ${error}`);
|
|
103
|
-
}
|
|
104
|
-
const data = await response.json();
|
|
105
|
-
const choice = data.choices[0];
|
|
106
|
-
return {
|
|
107
|
-
text: choice.message.content || undefined,
|
|
108
|
-
toolCalls: choice.message.tool_calls?.map(tc => ({
|
|
109
|
-
toolCallType: 'function',
|
|
110
|
-
toolCallId: tc.id,
|
|
111
|
-
toolName: tc.function.name,
|
|
112
|
-
args: tc.function.arguments,
|
|
113
|
-
})),
|
|
114
|
-
finishReason: choice.finish_reason || 'stop',
|
|
115
|
-
usage: {
|
|
116
|
-
promptTokens: data.usage?.prompt_tokens || 0,
|
|
117
|
-
completionTokens: data.usage?.completion_tokens || 0,
|
|
118
|
-
totalTokens: data.usage?.total_tokens || 0,
|
|
119
|
-
},
|
|
120
|
-
rawCall: {
|
|
121
|
-
rawPrompt: messages,
|
|
122
|
-
rawSettings: requestBody,
|
|
123
|
-
},
|
|
124
|
-
providerMetadata: data._meta ? {
|
|
125
|
-
a3m: {
|
|
126
|
-
provider: data._meta.provider,
|
|
127
|
-
cost: data._meta.cost,
|
|
128
|
-
},
|
|
129
|
-
} : undefined,
|
|
130
|
-
};
|
|
131
|
-
}
|
|
132
|
-
catch (error) {
|
|
133
|
-
throw new Error(`A3M Router generation failed: ${error instanceof Error ? error.message : String(error)}`);
|
|
134
|
-
}
|
|
135
|
-
},
|
|
136
|
-
doStream(options) {
|
|
137
|
-
const messages = convertMessages(options.prompt);
|
|
138
|
-
// Extract system prompt if present
|
|
139
|
-
const systemMessages = messages.filter(m => m.role === 'system');
|
|
140
|
-
const nonSystemMessages = messages.filter(m => m.role !== 'system');
|
|
141
|
-
const system = systemMessages.map(m => m.content).join('\n');
|
|
142
|
-
const requestBody = {
|
|
143
|
-
model: 'auto',
|
|
144
|
-
messages: nonSystemMessages.length > 0 ? nonSystemMessages : [{ role: 'user', content: ' ' }],
|
|
145
|
-
stream: true,
|
|
146
|
-
};
|
|
147
|
-
if (system) {
|
|
148
|
-
requestBody.system = system;
|
|
149
|
-
}
|
|
150
|
-
if (options.temperature !== undefined) {
|
|
151
|
-
requestBody.temperature = options.temperature;
|
|
152
|
-
}
|
|
153
|
-
if (options.maxTokens !== undefined) {
|
|
154
|
-
requestBody.max_tokens = options.maxTokens;
|
|
155
|
-
}
|
|
156
|
-
if (config.parallelEnsemble) {
|
|
157
|
-
requestBody.parallel = config.parallelCount || 3;
|
|
158
|
-
}
|
|
159
|
-
if (config.stealth) {
|
|
160
|
-
requestBody.stealth = true;
|
|
161
|
-
}
|
|
162
|
-
let attemptCount = 0;
|
|
163
|
-
const maxAttempts = 3;
|
|
164
|
-
const doStream = async () => {
|
|
165
|
-
attemptCount++;
|
|
166
|
-
try {
|
|
167
|
-
const response = await fetch(`${baseURL}/v1/chat/completions`, {
|
|
168
|
-
method: 'POST',
|
|
169
|
-
headers: {
|
|
170
|
-
'Content-Type': 'application/json',
|
|
171
|
-
'Authorization': `Bearer ${apiKey}`,
|
|
172
|
-
},
|
|
173
|
-
body: JSON.stringify(requestBody),
|
|
174
|
-
});
|
|
175
|
-
if (!response.ok) {
|
|
176
|
-
const error = await response.text();
|
|
177
|
-
throw new Error(`A3M Router streaming error: ${response.status} ${error}`);
|
|
178
|
-
}
|
|
179
|
-
if (!response.body) {
|
|
180
|
-
throw new Error('A3M Router: no response body');
|
|
181
|
-
}
|
|
182
|
-
const stream = new ReadableStream({
|
|
183
|
-
async start(controller) {
|
|
184
|
-
const reader = response.body.getReader();
|
|
185
|
-
const decoder = new TextDecoder();
|
|
186
|
-
let buffer = '';
|
|
187
|
-
try {
|
|
188
|
-
while (true) {
|
|
189
|
-
const { done, value } = await reader.read();
|
|
190
|
-
if (done) {
|
|
191
|
-
controller.close();
|
|
192
|
-
break;
|
|
193
|
-
}
|
|
194
|
-
buffer += decoder.decode(value, { stream: true });
|
|
195
|
-
const lines = buffer.split('\n');
|
|
196
|
-
buffer = lines.pop() || '';
|
|
197
|
-
for (const line of lines) {
|
|
198
|
-
if (line.startsWith('data: ')) {
|
|
199
|
-
const data = line.slice(6);
|
|
200
|
-
if (data === '[DONE]') {
|
|
201
|
-
controller.close();
|
|
202
|
-
return;
|
|
203
|
-
}
|
|
204
|
-
try {
|
|
205
|
-
const parsed = JSON.parse(data);
|
|
206
|
-
// Handle chat completion chunks
|
|
207
|
-
if (parsed.choices?.[0]?.delta?.content) {
|
|
208
|
-
const chunk = parsed.choices[0].delta.content;
|
|
209
|
-
controller.enqueue({
|
|
210
|
-
type: 'text-delta',
|
|
211
|
-
textDelta: chunk,
|
|
212
|
-
});
|
|
213
|
-
}
|
|
214
|
-
// Handle tool call start
|
|
215
|
-
if (parsed.choices?.[0]?.delta?.tool_calls?.[0]) {
|
|
216
|
-
const toolCall = parsed.choices[0].delta.tool_calls[0];
|
|
217
|
-
if (toolCall.id) {
|
|
218
|
-
controller.enqueue({
|
|
219
|
-
type: 'tool-call',
|
|
220
|
-
toolCallType: 'function',
|
|
221
|
-
toolCallId: toolCall.id,
|
|
222
|
-
toolName: toolCall.function?.name || '',
|
|
223
|
-
args: toolCall.function?.arguments || '',
|
|
224
|
-
});
|
|
225
|
-
}
|
|
226
|
-
}
|
|
227
|
-
// Handle tool call delta
|
|
228
|
-
if (parsed.choices?.[0]?.delta?.tool_calls?.[0]?.function?.arguments) {
|
|
229
|
-
const argsDelta = parsed.choices[0].delta.tool_calls[0].function.arguments;
|
|
230
|
-
if (argsDelta) {
|
|
231
|
-
controller.enqueue({
|
|
232
|
-
type: 'tool-call-delta',
|
|
233
|
-
toolCallType: 'function',
|
|
234
|
-
toolCallId: parsed.choices[0].delta.tool_calls[0].id,
|
|
235
|
-
toolName: parsed.choices[0].delta.tool_calls[0].function?.name || '',
|
|
236
|
-
argsTextDelta: argsDelta,
|
|
237
|
-
});
|
|
238
|
-
}
|
|
239
|
-
}
|
|
240
|
-
// Handle finish
|
|
241
|
-
if (parsed.choices?.[0]?.finish_reason) {
|
|
242
|
-
controller.enqueue({
|
|
243
|
-
type: 'finish',
|
|
244
|
-
finishReason: parsed.choices[0].finish_reason,
|
|
245
|
-
usage: {
|
|
246
|
-
promptTokens: parsed.usage?.prompt_tokens || 0,
|
|
247
|
-
completionTokens: parsed.usage?.completion_tokens || 0,
|
|
248
|
-
},
|
|
249
|
-
});
|
|
250
|
-
}
|
|
251
|
-
}
|
|
252
|
-
catch {
|
|
253
|
-
// Skip malformed JSON
|
|
254
|
-
}
|
|
255
|
-
}
|
|
256
|
-
}
|
|
257
|
-
}
|
|
258
|
-
}
|
|
259
|
-
catch (streamError) {
|
|
260
|
-
controller.error(streamError);
|
|
261
|
-
}
|
|
262
|
-
},
|
|
263
|
-
});
|
|
264
|
-
return {
|
|
265
|
-
stream,
|
|
266
|
-
rawCall: {
|
|
267
|
-
rawPrompt: messages,
|
|
268
|
-
rawSettings: requestBody,
|
|
269
|
-
},
|
|
270
|
-
};
|
|
271
|
-
}
|
|
272
|
-
catch (error) {
|
|
273
|
-
if (attemptCount < maxAttempts) {
|
|
274
|
-
await new Promise(resolve => setTimeout(resolve, 1000 * attemptCount));
|
|
275
|
-
return doStream();
|
|
276
|
-
}
|
|
277
|
-
throw error;
|
|
278
|
-
}
|
|
279
|
-
};
|
|
280
|
-
return new Promise((resolve, reject) => {
|
|
281
|
-
doStream()
|
|
282
|
-
.then(resolve)
|
|
283
|
-
.catch(reject);
|
|
284
|
-
});
|
|
285
|
-
},
|
|
286
|
-
};
|
|
287
|
-
return model;
|
|
288
|
-
}
|
|
289
|
-
//# sourceMappingURL=a3m-language-model.js.map
|
|
@@ -1 +0,0 @@
|
|
|
1
|
-
{"version":3,"file":"a3m-language-model.js","sourceRoot":"","sources":["../src/a3m-language-model.ts"],"names":[],"mappings":"AAAA;;;;GAIG;AAaH;;GAEG;AACH,SAAS,eAAe,CACtB,MAA4C;IAE5C,MAAM,MAAM,GAAsE,EAAE,CAAC;IAErF,wBAAwB;IACxB,IAAI,OAAO,MAAM,KAAK,QAAQ,EAAE,CAAC;QAC/B,OAAO,CAAC,EAAE,IAAI,EAAE,MAAM,EAAE,OAAO,EAAE,MAAM,EAAE,CAAC,CAAC;IAC7C,CAAC;IAED,uCAAuC;IACvC,KAAK,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;QAC1B,IAAI,IAAI,CAAC,IAAI,KAAK,QAAQ,EAAE,CAAC;YAC3B,MAAM,CAAC,IAAI,CAAC,EAAE,IAAI,EAAE,QAAQ,EAAE,OAAO,EAAE,IAAI,CAAC,OAAO,EAAE,CAAC,CAAC;QACzD,CAAC;aAAM,IAAI,IAAI,CAAC,IAAI,KAAK,MAAM,EAAE,CAAC;YAChC,IAAI,OAAO,IAAI,CAAC,OAAO,KAAK,QAAQ,EAAE,CAAC;gBACrC,MAAM,CAAC,IAAI,CAAC,EAAE,IAAI,EAAE,MAAM,EAAE,OAAO,EAAE,IAAI,CAAC,OAAO,EAAE,CAAC,CAAC;YACvD,CAAC;iBAAM,IAAI,KAAK,CAAC,OAAO,CAAC,IAAI,CAAC,OAAO,CAAC,EAAE,CAAC;gBACvC,MAAM,WAAW,GAAG,IAAI,CAAC,OAAO;qBAC7B,MAAM,CAAC,CAAC,CAAC,EAAuC,EAAE,CAAC,CAAC,CAAC,IAAI,KAAK,MAAM,CAAC;qBACrE,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC;qBAChB,IAAI,CAAC,IAAI,CAAC,CAAC;gBACd,MAAM,CAAC,IAAI,CAAC,EAAE,IAAI,EAAE,MAAM,EAAE,OAAO,EAAE,WAAW,EAAE,CAAC,CAAC;YACtD,CAAC;QACH,CAAC;aAAM,IAAI,IAAI,CAAC,IAAI,KAAK,WAAW,EAAE,CAAC;YACrC,IAAI,OAAO,IAAI,CAAC,OAAO,KAAK,QAAQ,EAAE,CAAC;gBACrC,MAAM,CAAC,IAAI,CAAC,EAAE,IAAI,EAAE,WAAW,EAAE,OAAO,EAAE,IAAI,CAAC,OAAO,EAAE,CAAC,CAAC;YAC5D,CAAC;iBAAM,IAAI,KAAK,CAAC,OAAO,CAAC,IAAI,CAAC,OAAO,CAAC,EAAE,CAAC;gBACvC,MAAM,WAAW,GAAG,IAAI,CAAC,OAAO;qBAC7B,MAAM,CAAC,CAAC,CAAC,EAAuC,EAAE,CAAC,CAAC,CAAC,IAAI,KAAK,MAAM,CAAC;qBACrE,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC;qBAChB,IAAI,CAAC,IAAI,CAAC,CAAC;gBACd,MAAM,CAAC,IAAI,CAAC,EAAE,IAAI,EAAE,WAAW,EAAE,OAAO,EAAE,WAAW,EAAE,CAAC,CAAC;YAC3D,CAAC;QACH,CAAC;IACH,CAAC;IAED,OAAO,MAAM,CAAC;AAChB,CAAC;AAED;;GAEG;AACH,MAAM,UAAU,sBAAsB,CAAC,SAA0B,EAAE;IACjE,MAAM,OAAO,GAAG,MAAM,CAAC,OAAO,IAAI,uBAAuB,CAAC;IAC1D,MAAM,MAAM,GAAG,MAAM,CAAC,MAAM,IAAI,YAAY,CAAC;IAE7C,MAAM,KAAK,GAAoB;QAC7B,oBAAoB,EAAE,IAAI;QAC1B,QAAQ,EAAE,KAAK;QACf,OAAO,EAAE,MAAM;QACf,2BAA2B,EAAE,SAAS;QAEtC,KAAK,CAAC,UAAU,CAAC,OAAmC;YAgBlD,MAAM,QAAQ,GAAG,eAAe,CAAC,OAAO,CAAC,MAAM,CAAC,CAAC;YAEjD,mCAAmC;YACnC,MAAM,cAAc,GAAG,QAAQ,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,IAAI,KAAK,QAAQ,CAAC,CAAC;YACjE,MAAM,iBAAiB,GAAG,QAAQ,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,IAAI,KAAK,QAAQ,CAAC,CAAC;YACpE,MAAM,MAAM,GAAG,cAAc,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;YAE7D,iCAAiC;YACjC,MAAM,WAAW,GAA4B;gBAC3C,KAAK,EAAE,MAAM;gBACb,QAAQ,EAAE,iBAAiB,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,CAAC,iBAAiB,CAAC,CAAC,CAAC,CAAC,EAAE,IAAI,EAAE,MAAM,EAAE,OAAO,EAAE,GAAG,EAAE,CAAC;aAC9F,CAAC;YAEF,IAAI,MAAM,EAAE,CAAC;gBACX,WAAW,CAAC,MAAM,GAAG,MAAM,CAAC;YAC9B,CAAC;YAED,IAAI,OAAO,CAAC,WAAW,KAAK,SAAS,EAAE,CAAC;gBACtC,WAAW,CAAC,WAAW,GAAG,OAAO,CAAC,WAAW,CAAC;YAChD,CAAC;YAED,IAAI,OAAO,CAAC,SAAS,KAAK,SAAS,EAAE,CAAC;gBACpC,WAAW,CAAC,UAAU,GAAG,OAAO,CAAC,SAAS,CAAC;YAC7C,CAAC;YAED,IAAI,OAAO,CAAC,IAAI,KAAK,SAAS,EAAE,CAAC;gBAC/B,WAAW,CAAC,KAAK,GAAG,OAAO,CAAC,IAAI,CAAC;YACnC,CAAC;YAED,IAAI,OAAO,CAAC,aAAa,KAAK,SAAS,EAAE,CAAC;gBACxC,WAAW,CAAC,IAAI,GAAG,OAAO,CAAC,aAAa,CAAC;YAC3C,CAAC;YAED,IAAI,MAAM,CAAC,gBAAgB,EAAE,CAAC;gBAC5B,WAAW,CAAC,QAAQ,GAAG,MAAM,CAAC,aAAa,IAAI,CAAC,CAAC;YACnD,CAAC;YAED,IAAI,MAAM,CAAC,OAAO,EAAE,CAAC;gBACnB,WAAW,CAAC,OAAO,GAAG,IAAI,CAAC;YAC7B,CAAC;YAED,IAAI,CAAC;gBACH,kBAAkB;gBAClB,MAAM,QAAQ,GAAG,MAAM,KAAK,CAAC,GAAG,OAAO,sBAAsB,EAAE;oBAC7D,MAAM,EAAE,MAAM;oBACd,OAAO,EAAE;wBACP,cAAc,EAAE,kBAAkB;wBAClC,eAAe,EAAE,UAAU,MAAM,EAAE;qBACpC;oBACD,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC,WAAW,CAAC;iBAClC,CAAC,CAAC;gBAEH,IAAI,CAAC,QAAQ,CAAC,EAAE,EAAE,CAAC;oBACjB,MAAM,KAAK,GAAG,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC;oBACpC,MAAM,IAAI,KAAK,CAAC,qBAAqB,QAAQ,CAAC,MAAM,IAAI,KAAK,EAAE,CAAC,CAAC;gBACnE,CAAC;gBAED,MAAM,IAAI,GAAG,MAAM,QAAQ,CAAC,IAAI,EAoB/B,CAAC;gBAEF,MAAM,MAAM,GAAG,IAAI,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC;gBAE/B,OAAO;oBACL,IAAI,EAAE,MAAM,CAAC,OAAO,CAAC,OAAO,IAAI,SAAS;oBACzC,SAAS,EAAE,MAAM,CAAC,OAAO,CAAC,UAAU,EAAE,GAAG,CAAC,EAAE,CAAC,EAAE,CAAC,CAAC;wBAC/C,YAAY,EAAE,UAAmB;wBACjC,UAAU,EAAE,EAAE,CAAC,EAAE;wBACjB,QAAQ,EAAE,EAAE,CAAC,QAAQ,CAAC,IAAI;wBAC1B,IAAI,EAAE,EAAE,CAAC,QAAQ,CAAC,SAAS;qBAC5B,CAAC,CAAC;oBACH,YAAY,EAAG,MAAM,CAAC,aAA6C,IAAI,MAAM;oBAC7E,KAAK,EAAE;wBACL,YAAY,EAAE,IAAI,CAAC,KAAK,EAAE,aAAa,IAAI,CAAC;wBAC5C,gBAAgB,EAAE,IAAI,CAAC,KAAK,EAAE,iBAAiB,IAAI,CAAC;wBACpD,WAAW,EAAE,IAAI,CAAC,KAAK,EAAE,YAAY,IAAI,CAAC;qBAC3C;oBACD,OAAO,EAAE;wBACP,SAAS,EAAE,QAAQ;wBACnB,WAAW,EAAE,WAAW;qBACzB;oBACD,gBAAgB,EAAE,IAAI,CAAC,KAAK,CAAC,CAAC,CAAC;wBAC7B,GAAG,EAAE;4BACH,QAAQ,EAAE,IAAI,CAAC,KAAK,CAAC,QAAQ;4BAC7B,IAAI,EAAE,IAAI,CAAC,KAAK,CAAC,IAAI;yBACtB;qBACF,CAAC,CAAC,CAAC,SAAS;iBACd,CAAC;YACJ,CAAC;YAAC,OAAO,KAAK,EAAE,CAAC;gBACf,MAAM,IAAI,KAAK,CACb,iCAAiC,KAAK,YAAY,KAAK,CAAC,CAAC,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC,CAAC,MAAM,CAAC,KAAK,CAAC,EAAE,CAC1F,CAAC;YACJ,CAAC;QACH,CAAC;QAED,QAAQ,CAAC,OAAmC;YAC1C,MAAM,QAAQ,GAAG,eAAe,CAAC,OAAO,CAAC,MAAM,CAAC,CAAC;YAEjD,mCAAmC;YACnC,MAAM,cAAc,GAAG,QAAQ,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,IAAI,KAAK,QAAQ,CAAC,CAAC;YACjE,MAAM,iBAAiB,GAAG,QAAQ,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,IAAI,KAAK,QAAQ,CAAC,CAAC;YACpE,MAAM,MAAM,GAAG,cAAc,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;YAE7D,MAAM,WAAW,GAA4B;gBAC3C,KAAK,EAAE,MAAM;gBACb,QAAQ,EAAE,iBAAiB,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,CAAC,iBAAiB,CAAC,CAAC,CAAC,CAAC,EAAE,IAAI,EAAE,MAAM,EAAE,OAAO,EAAE,GAAG,EAAE,CAAC;gBAC7F,MAAM,EAAE,IAAI;aACb,CAAC;YAEF,IAAI,MAAM,EAAE,CAAC;gBACX,WAAW,CAAC,MAAM,GAAG,MAAM,CAAC;YAC9B,CAAC;YAED,IAAI,OAAO,CAAC,WAAW,KAAK,SAAS,EAAE,CAAC;gBACtC,WAAW,CAAC,WAAW,GAAG,OAAO,CAAC,WAAW,CAAC;YAChD,CAAC;YAED,IAAI,OAAO,CAAC,SAAS,KAAK,SAAS,EAAE,CAAC;gBACpC,WAAW,CAAC,UAAU,GAAG,OAAO,CAAC,SAAS,CAAC;YAC7C,CAAC;YAED,IAAI,MAAM,CAAC,gBAAgB,EAAE,CAAC;gBAC5B,WAAW,CAAC,QAAQ,GAAG,MAAM,CAAC,aAAa,IAAI,CAAC,CAAC;YACnD,CAAC;YAED,IAAI,MAAM,CAAC,OAAO,EAAE,CAAC;gBACnB,WAAW,CAAC,OAAO,GAAG,IAAI,CAAC;YAC7B,CAAC;YAED,IAAI,YAAY,GAAG,CAAC,CAAC;YACrB,MAAM,WAAW,GAAG,CAAC,CAAC;YAEtB,MAAM,QAAQ,GAAG,KAAK,IAGnB,EAAE;gBACH,YAAY,EAAE,CAAC;gBAEf,IAAI,CAAC;oBACH,MAAM,QAAQ,GAAG,MAAM,KAAK,CAAC,GAAG,OAAO,sBAAsB,EAAE;wBAC7D,MAAM,EAAE,MAAM;wBACd,OAAO,EAAE;4BACP,cAAc,EAAE,kBAAkB;4BAClC,eAAe,EAAE,UAAU,MAAM,EAAE;yBACpC;wBACD,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC,WAAW,CAAC;qBAClC,CAAC,CAAC;oBAEH,IAAI,CAAC,QAAQ,CAAC,EAAE,EAAE,CAAC;wBACjB,MAAM,KAAK,GAAG,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC;wBACpC,MAAM,IAAI,KAAK,CAAC,+BAA+B,QAAQ,CAAC,MAAM,IAAI,KAAK,EAAE,CAAC,CAAC;oBAC7E,CAAC;oBAED,IAAI,CAAC,QAAQ,CAAC,IAAI,EAAE,CAAC;wBACnB,MAAM,IAAI,KAAK,CAAC,8BAA8B,CAAC,CAAC;oBAClD,CAAC;oBAED,MAAM,MAAM,GAAG,IAAI,cAAc,CAA4B;wBAC3D,KAAK,CAAC,KAAK,CAAC,UAAU;4BACpB,MAAM,MAAM,GAAG,QAAQ,CAAC,IAAK,CAAC,SAAS,EAAE,CAAC;4BAC1C,MAAM,OAAO,GAAG,IAAI,WAAW,EAAE,CAAC;4BAClC,IAAI,MAAM,GAAG,EAAE,CAAC;4BAEhB,IAAI,CAAC;gCACH,OAAO,IAAI,EAAE,CAAC;oCACZ,MAAM,EAAE,IAAI,EAAE,KAAK,EAAE,GAAG,MAAM,MAAM,CAAC,IAAI,EAAE,CAAC;oCAE5C,IAAI,IAAI,EAAE,CAAC;wCACT,UAAU,CAAC,KAAK,EAAE,CAAC;wCACnB,MAAM;oCACR,CAAC;oCAED,MAAM,IAAI,OAAO,CAAC,MAAM,CAAC,KAAK,EAAE,EAAE,MAAM,EAAE,IAAI,EAAE,CAAC,CAAC;oCAClD,MAAM,KAAK,GAAG,MAAM,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC;oCACjC,MAAM,GAAG,KAAK,CAAC,GAAG,EAAE,IAAI,EAAE,CAAC;oCAE3B,KAAK,MAAM,IAAI,IAAI,KAAK,EAAE,CAAC;wCACzB,IAAI,IAAI,CAAC,UAAU,CAAC,QAAQ,CAAC,EAAE,CAAC;4CAC9B,MAAM,IAAI,GAAG,IAAI,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;4CAE3B,IAAI,IAAI,KAAK,QAAQ,EAAE,CAAC;gDACtB,UAAU,CAAC,KAAK,EAAE,CAAC;gDACnB,OAAO;4CACT,CAAC;4CAED,IAAI,CAAC;gDACH,MAAM,MAAM,GAAG,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC;gDAEhC,gCAAgC;gDAChC,IAAI,MAAM,CAAC,OAAO,EAAE,CAAC,CAAC,CAAC,EAAE,KAAK,EAAE,OAAO,EAAE,CAAC;oDACxC,MAAM,KAAK,GAAG,MAAM,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC,OAAO,CAAC;oDAC9C,UAAU,CAAC,OAAO,CAAC;wDACjB,IAAI,EAAE,YAAY;wDAClB,SAAS,EAAE,KAAK;qDACY,CAAC,CAAC;gDAClC,CAAC;gDAED,yBAAyB;gDACzB,IAAI,MAAM,CAAC,OAAO,EAAE,CAAC,CAAC,CAAC,EAAE,KAAK,EAAE,UAAU,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;oDAChD,MAAM,QAAQ,GAAG,MAAM,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC,UAAU,CAAC,CAAC,CAAC,CAAC;oDACvD,IAAI,QAAQ,CAAC,EAAE,EAAE,CAAC;wDAChB,UAAU,CAAC,OAAO,CAAC;4DACjB,IAAI,EAAE,WAAW;4DACjB,YAAY,EAAE,UAAU;4DACxB,UAAU,EAAE,QAAQ,CAAC,EAAE;4DACvB,QAAQ,EAAE,QAAQ,CAAC,QAAQ,EAAE,IAAI,IAAI,EAAE;4DACvC,IAAI,EAAE,QAAQ,CAAC,QAAQ,EAAE,SAAS,IAAI,EAAE;yDACZ,CAAC,CAAC;oDAClC,CAAC;gDACH,CAAC;gDAED,yBAAyB;gDACzB,IAAI,MAAM,CAAC,OAAO,EAAE,CAAC,CAAC,CAAC,EAAE,KAAK,EAAE,UAAU,EAAE,CAAC,CAAC,CAAC,EAAE,QAAQ,EAAE,SAAS,EAAE,CAAC;oDACrE,MAAM,SAAS,GAAG,MAAM,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC,UAAU,CAAC,CAAC,CAAC,CAAC,QAAQ,CAAC,SAAS,CAAC;oDAC3E,IAAI,SAAS,EAAE,CAAC;wDACd,UAAU,CAAC,OAAO,CAAC;4DACjB,IAAI,EAAE,iBAAiB;4DACvB,YAAY,EAAE,UAAU;4DACxB,UAAU,EAAE,MAAM,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC,UAAU,CAAC,CAAC,CAAC,CAAC,EAAE;4DACpD,QAAQ,EAAE,MAAM,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC,UAAU,CAAC,CAAC,CAAC,CAAC,QAAQ,EAAE,IAAI,IAAI,EAAE;4DACpE,aAAa,EAAE,SAAS;yDACI,CAAC,CAAC;oDAClC,CAAC;gDACH,CAAC;gDAED,gBAAgB;gDAChB,IAAI,MAAM,CAAC,OAAO,EAAE,CAAC,CAAC,CAAC,EAAE,aAAa,EAAE,CAAC;oDACvC,UAAU,CAAC,OAAO,CAAC;wDACjB,IAAI,EAAE,QAAQ;wDACd,YAAY,EAAE,MAAM,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,aAA4C;wDAC5E,KAAK,EAAE;4DACL,YAAY,EAAE,MAAM,CAAC,KAAK,EAAE,aAAa,IAAI,CAAC;4DAC9C,gBAAgB,EAAE,MAAM,CAAC,KAAK,EAAE,iBAAiB,IAAI,CAAC;yDACvD;qDAC2B,CAAC,CAAC;gDAClC,CAAC;4CACH,CAAC;4CAAC,MAAM,CAAC;gDACP,sBAAsB;4CACxB,CAAC;wCACH,CAAC;oCACH,CAAC;gCACH,CAAC;4BACH,CAAC;4BAAC,OAAO,WAAW,EAAE,CAAC;gCACrB,UAAU,CAAC,KAAK,CAAC,WAAW,CAAC,CAAC;4BAChC,CAAC;wBACH,CAAC;qBACF,CAAC,CAAC;oBAEH,OAAO;wBACL,MAAM;wBACN,OAAO,EAAE;4BACP,SAAS,EAAE,QAAQ;4BACnB,WAAW,EAAE,WAAW;yBACzB;qBACF,CAAC;gBACJ,CAAC;gBAAC,OAAO,KAAK,EAAE,CAAC;oBACf,IAAI,YAAY,GAAG,WAAW,EAAE,CAAC;wBAC/B,MAAM,IAAI,OAAO,CAAC,OAAO,CAAC,EAAE,CAAC,UAAU,CAAC,OAAO,EAAE,IAAI,GAAG,YAAY,CAAC,CAAC,CAAC;wBACvE,OAAO,QAAQ,EAAE,CAAC;oBACpB,CAAC;oBACD,MAAM,KAAK,CAAC;gBACd,CAAC;YACH,CAAC,CAAC;YAEF,OAAO,IAAI,OAAO,CAAC,CAAC,OAAO,EAAE,MAAM,EAAE,EAAE;gBACrC,QAAQ,EAAE;qBACP,IAAI,CAAC,OAAO,CAAC;qBACb,KAAK,CAAC,MAAM,CAAC,CAAC;YACnB,CAAC,CAAC,CAAC;QACL,CAAC;KACF,CAAC;IAEF,OAAO,KAAK,CAAC;AACf,CAAC"}
|
|
@@ -1,82 +0,0 @@
|
|
|
1
|
-
/**
|
|
2
|
-
* @a3m/vercel-ai - A3M Router Provider for Vercel AI SDK
|
|
3
|
-
*
|
|
4
|
-
* Drop-in replacement that routes to the cheapest capable provider
|
|
5
|
-
* with parallel execution, caching, and automatic fallback.
|
|
6
|
-
*
|
|
7
|
-
* @example
|
|
8
|
-
* ```typescript
|
|
9
|
-
* import { createA3MProvider } from '@a3m/vercel-ai';
|
|
10
|
-
* import { generateText } from 'ai';
|
|
11
|
-
*
|
|
12
|
-
* const a3m = createA3MProvider();
|
|
13
|
-
*
|
|
14
|
-
* const { text } = await generateText({
|
|
15
|
-
* model: a3m('auto'),
|
|
16
|
-
* prompt: 'What is the capital of France?',
|
|
17
|
-
* });
|
|
18
|
-
* ```
|
|
19
|
-
*/
|
|
20
|
-
export type { A3MRouterConfig, A3MRouteResponse, A3MStreamChunk } from './types';
|
|
21
|
-
export { createA3MLanguageModel } from './a3m-language-model';
|
|
22
|
-
import type { LanguageModelV1 } from '@ai-sdk/provider';
|
|
23
|
-
import type { A3MRouterConfig } from './types';
|
|
24
|
-
/**
|
|
25
|
-
* A3M Router provider for Vercel AI SDK
|
|
26
|
-
*/
|
|
27
|
-
export interface A3MProvider {
|
|
28
|
-
name: 'a3m';
|
|
29
|
-
/**
|
|
30
|
-
* Returns an A3M language model.
|
|
31
|
-
*
|
|
32
|
-
* @param modelId - Use 'auto' for automatic routing, or specify a model like 'gpt-4', 'claude-3', etc.
|
|
33
|
-
*/
|
|
34
|
-
(modelId: string): LanguageModelV1;
|
|
35
|
-
}
|
|
36
|
-
/**
|
|
37
|
-
* Create an A3M Router provider for Vercel AI SDK
|
|
38
|
-
*/
|
|
39
|
-
export declare function createA3MProvider(config?: A3MRouterConfig): A3MProvider;
|
|
40
|
-
/**
|
|
41
|
-
* Example usage with Next.js App Router
|
|
42
|
-
*
|
|
43
|
-
* ```typescript
|
|
44
|
-
* // app/api/chat/route.ts
|
|
45
|
-
* import { createA3MProvider } from '@a3m/vercel-ai';
|
|
46
|
-
* import { generateText } from 'ai';
|
|
47
|
-
*
|
|
48
|
-
* const a3m = createA3MProvider({
|
|
49
|
-
* parallelEnsemble: true,
|
|
50
|
-
* });
|
|
51
|
-
*
|
|
52
|
-
* export async function POST(req: Request) {
|
|
53
|
-
* const { messages } = await req.json();
|
|
54
|
-
*
|
|
55
|
-
* const result = await generateText({
|
|
56
|
-
* model: a3m('auto'),
|
|
57
|
-
* messages,
|
|
58
|
-
* });
|
|
59
|
-
*
|
|
60
|
-
* return Response.json(result);
|
|
61
|
-
* }
|
|
62
|
-
* ```
|
|
63
|
-
*/
|
|
64
|
-
/**
|
|
65
|
-
* Example usage with streaming
|
|
66
|
-
*
|
|
67
|
-
* ```typescript
|
|
68
|
-
* import { createA3MProvider } from '@a3m/vercel-ai';
|
|
69
|
-
* import { streamText } from 'ai';
|
|
70
|
-
*
|
|
71
|
-
* const a3m = createA3MProvider();
|
|
72
|
-
*
|
|
73
|
-
* const result = await streamText({
|
|
74
|
-
* model: a3m('auto'),
|
|
75
|
-
* prompt: 'Write a story about a robot...',
|
|
76
|
-
* });
|
|
77
|
-
*
|
|
78
|
-
* // Stream to response
|
|
79
|
-
* return result.toDataStreamResponse();
|
|
80
|
-
* ```
|
|
81
|
-
*/
|
|
82
|
-
//# sourceMappingURL=index.d.ts.map
|
|
@@ -1 +0,0 @@
|
|
|
1
|
-
{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;GAkBG;AAEH,YAAY,EAAE,eAAe,EAAE,gBAAgB,EAAE,cAAc,EAAE,MAAM,SAAS,CAAC;AACjF,OAAO,EAAE,sBAAsB,EAAE,MAAM,sBAAsB,CAAC;AAE9D,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,kBAAkB,CAAC;AACxD,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,SAAS,CAAC;AAG/C;;GAEG;AACH,MAAM,WAAW,WAAW;IAC1B,IAAI,EAAE,KAAK,CAAC;IACZ;;;;OAIG;IACH,CAAC,OAAO,EAAE,MAAM,GAAG,eAAe,CAAC;CACpC;AAED;;GAEG;AACH,wBAAgB,iBAAiB,CAAC,MAAM,GAAE,eAAoB,GAAG,WAAW,CAgB3E;AAED;;;;;;;;;;;;;;;;;;;;;;;GAuBG;AAEH;;;;;;;;;;;;;;;;;GAiBG"}
|