@orquentra/codebuddy 2.5.0 → 2.6.0

package/README.md

@@ -20,6 +20,14 @@ The setup wizard configures the project and agent integration. Project context
 works without a database; semantic memory recall optionally uses PostgreSQL
 with pgvector and a local Hugging Face embedding model (no token needed).
 
+Run setup inside the project you want CodeBuddy to read. Claude Desktop
+registration saves that folder's absolute path, so restarting the desktop app
+does not change which project is indexed. `serve` requires an initialized
+project (`.codebuddy/config.json`) and refuses filesystem roots and home folders.
+Older desktop registrations without a project path can be repaired by running
+`codebuddy claude install` from the initialized project folder, then restarting
+Claude Desktop.
+
 ## What it does
 
 - **Task context:** relevant files, symbols, code excerpts, nearby dependencies,

package/dist/{chunk-KYK6RPAJ.js → chunk-JFDRBPYF.js}

@@ -1,6 +1,6 @@
 import {
   EMBEDDING_DIMENSIONS
-} from "./chunk-JMWJ7V6Y.js";
+} from "./chunk-JPHJBCZB.js";
 
 // src/providers/cloud.ts
 var OPENROUTER_BASE_URL = "https://openrouter.ai/api/v1";
@@ -96,4 +96,4 @@ export {
   rerankOpenRouter,
   rerankHuggingFace
 };
-//# sourceMappingURL=chunk-KYK6RPAJ.js.map
+//# sourceMappingURL=chunk-JFDRBPYF.js.map

package/dist/{chunk-JMWJ7V6Y.js → chunk-JPHJBCZB.js}

@@ -8,7 +8,7 @@ var LOCAL_EMBEDDING_MODEL = {
   revision: "1dc7835ba0cb9c76a3618d0bf0c427c97671b3c8",
   schema: "transformers.js@4-q8-cls-l2",
   supportsBatch: true,
-  relevanceFloor: 0.81
+  relevanceFloor: 0.84
 };
 var DEFAULT_RERANK_MODELS = {
   local: {
@@ -96,4 +96,4 @@ export {
   getModelContextWindow,
   assertCompatibleEmbeddingModels
 };
-//# sourceMappingURL=chunk-JMWJ7V6Y.js.map
+//# sourceMappingURL=chunk-JPHJBCZB.js.map

package/dist/{chunk-JMWJ7V6Y.js.map → chunk-JPHJBCZB.js.map}

@@ -1 +1 @@
-{"version":3,"sources":["../src/providers/models.ts"],"sourcesContent":["export type ModelRole = \"primary\" | \"fallback\" | \"last_resort\";\nexport type ModelKind = \"embedding\" | \"llm\";\n\nexport type ModelDefinition = {\n  id: string;\n  role: ModelRole;\n  kind: ModelKind;\n  dimensions?: number;\n  /** Actual deployment revision metadata; this does not pin Hugging Face requests. */\n  revision?: string;\n  schema?: string;\n  contextWindow?: number;\n  dailyCallBudget?: number;\n  supportsBatch?: boolean;\n  /**\n   * Cosine at or above which a memory counts as semantically relevant to a\n   * task on its own (without a shared word or path). Calibrated per model by\n   * `npm run eval:semantic`; cosines are not comparable across models.\n   */\n  relevanceFloor?: number;\n};\n\nexport const EMBEDDING_DIMENSIONS = 384;\n\n/**\n * Default local embedder (ONNX via transformers.js, no network after the first\n * download): ModernBERT-based, 384-d, 8k context, trained on code retrieval.\n * The revision is a pinned commit so the embedding space never drifts.\n */\nexport const LOCAL_EMBEDDING_MODEL = {\n  id: \"onnx-community/granite-embedding-small-english-r2-ONNX\",\n  role: \"primary\",\n  kind: \"embedding\",\n  dimensions: EMBEDDING_DIMENSIONS,\n  revision: \"1dc7835ba0cb9c76a3618d0bf0c427c97671b3c8\",\n  schema: \"transformers.js@4-q8-cls-l2\",\n  supportsBatch: true,\n  relevanceFloor: 0.81,\n} satisfies ModelDefinition;\n\n/**\n * Cross-encoder rerankers per `rerank.provider` (off by default: on\n * `npm run eval:semantic` none beat granite's dense order). The local default\n * was the best small one measured (13/20 first results, ~20 ms per prompt).\n */\nexport const DEFAULT_RERANK_MODELS = {\n  local: {\n    id: \"mixedbread-ai/mxbai-rerank-xsmall-v1\",\n    revision: \"b5c6e9da73abc3711f593f705371cdbe9e0fe422\",\n  },\n  huggingface: { id: \"BAAI/bge-reranker-v2-m3\" },\n  openrouter: { id: \"cohere/rerank-v3.5\" },\n} as const;\n\n/** Default hosted embedders; each must return 384-d vectors. */\nexport const DEFAULT_CLOUD_EMBEDDING_MODELS = {\n  huggingface: \"sentence-transformers/all-MiniLM-L6-v2\",\n  openrouter: \"openai/text-embedding-3-small\",\n} as const;\n\nexport const DEFAULT_EMBEDDING_MODELS = [\n  {\n    id: \"sentence-transformers/all-MiniLM-L6-v2\",\n    role: \"primary\",\n    kind: \"embedding\",\n    dimensions: EMBEDDING_DIMENSIONS,\n    dailyCallBudget: 1_000,\n    supportsBatch: true,\n    relevanceFloor: 0.42,\n  },\n  {\n    id: \"BAAI/bge-small-en-v1.5\",\n    role: \"fallback\",\n    kind: \"embedding\",\n    dimensions: EMBEDDING_DIMENSIONS,\n    dailyCallBudget: 1_000,\n    supportsBatch: true,\n    relevanceFloor: 0.69,\n  },\n] satisfies ModelDefinition[];\n\nexport const DEFAULT_LLM_MODELS = [\n  {\n    id: \"meta-llama/Llama-3.2-3B-Instruct\",\n    role: \"primary\",\n    kind: \"llm\",\n    contextWindow: 131_072,\n    dailyCallBudget: 300,\n  },\n  {\n    id: \"Qwen/Qwen2.5-7B-Instruct\",\n    role: \"fallback\",\n    kind: \"llm\",\n    contextWindow: 32_768,\n    dailyCallBudget: 300,\n  },\n  {\n    id: \"mistralai/Mistral-7B-Instruct-v0.3\",\n    role: \"last_resort\",\n    kind: \"llm\",\n    contextWindow: 32_768,\n    dailyCallBudget: 300,\n  },\n] satisfies ModelDefinition[];\n\nexport const MODEL_REGISTRY: ModelDefinition[] = [\n  LOCAL_EMBEDDING_MODEL,\n  ...DEFAULT_EMBEDDING_MODELS,\n  ...DEFAULT_LLM_MODELS,\n];\n\nexport function getModelDefinition(modelId: string): ModelDefinition | undefined {\n  return MODEL_REGISTRY.find((model) => model.id === modelId);\n}\n\nexport function getModelContextWindow(modelId: string): number | undefined {\n  return getModelDefinition(modelId)?.contextWindow;\n}\n\nexport function assertCompatibleEmbeddingModels(models: ModelDefinition[]): void {\n  const mismatchedModel = models.find((model) => model.dimensions !== EMBEDDING_DIMENSIONS);\n\n  if (mismatchedModel) {\n    throw new Error(\n      `Embedding model ${mismatchedModel.id} has ${mismatchedModel.dimensions ?? \"unknown\"} dimensions; expected ${EMBEDDING_DIMENSIONS}.`,\n    );\n  }\n}\n"],"mappings":";AAsBO,IAAM,uBAAuB;AAO7B,IAAM,wBAAwB;AAAA,EACnC,IAAI;AAAA,EACJ,MAAM;AAAA,EACN,MAAM;AAAA,EACN,YAAY;AAAA,EACZ,UAAU;AAAA,EACV,QAAQ;AAAA,EACR,eAAe;AAAA,EACf,gBAAgB;AAClB;AAOO,IAAM,wBAAwB;AAAA,EACnC,OAAO;AAAA,IACL,IAAI;AAAA,IACJ,UAAU;AAAA,EACZ;AAAA,EACA,aAAa,EAAE,IAAI,0BAA0B;AAAA,EAC7C,YAAY,EAAE,IAAI,qBAAqB;AACzC;AAGO,IAAM,iCAAiC;AAAA,EAC5C,aAAa;AAAA,EACb,YAAY;AACd;AAEO,IAAM,2BAA2B;AAAA,EACtC;AAAA,IACE,IAAI;AAAA,IACJ,MAAM;AAAA,IACN,MAAM;AAAA,IACN,YAAY;AAAA,IACZ,iBAAiB;AAAA,IACjB,eAAe;AAAA,IACf,gBAAgB;AAAA,EAClB;AAAA,EACA;AAAA,IACE,IAAI;AAAA,IACJ,MAAM;AAAA,IACN,MAAM;AAAA,IACN,YAAY;AAAA,IACZ,iBAAiB;AAAA,IACjB,eAAe;AAAA,IACf,gBAAgB;AAAA,EAClB;AACF;AAEO,IAAM,qBAAqB;AAAA,EAChC;AAAA,IACE,IAAI;AAAA,IACJ,MAAM;AAAA,IACN,MAAM;AAAA,IACN,eAAe;AAAA,IACf,iBAAiB;AAAA,EACnB;AAAA,EACA;AAAA,IACE,IAAI;AAAA,IACJ,MAAM;AAAA,IACN,MAAM;AAAA,IACN,eAAe;AAAA,IACf,iBAAiB;AAAA,EACnB;AAAA,EACA;AAAA,IACE,IAAI;AAAA,IACJ,MAAM;AAAA,IACN,MAAM;AAAA,IACN,eAAe;AAAA,IACf,iBAAiB;AAAA,EACnB;AACF;AAEO,IAAM,iBAAoC;AAAA,EAC/C;AAAA,EACA,GAAG;AAAA,EACH,GAAG;AACL;AAEO,SAAS,mBAAmB,SAA8C;AAC/E,SAAO,eAAe,KAAK,CAAC,UAAU,MAAM,OAAO,OAAO;AAC5D;AAEO,SAAS,sBAAsB,SAAqC;AACzE,SAAO,mBAAmB,OAAO,GAAG;AACtC;AAEO,SAAS,gCAAgC,QAAiC;AAC/E,QAAM,kBAAkB,OAAO,KAAK,CAAC,UAAU,MAAM,eAAe,oBAAoB;AAExF,MAAI,iBAAiB;AACnB,UAAM,IAAI;AAAA,MACR,mBAAmB,gBAAgB,EAAE,QAAQ,gBAAgB,cAAc,SAAS,yBAAyB,oBAAoB;AAAA,IACnI;AAAA,EACF;AACF;","names":[]}
+{"version":3,"sources":["../src/providers/models.ts"],"sourcesContent":["export type ModelRole = \"primary\" | \"fallback\" | \"last_resort\";\nexport type ModelKind = \"embedding\" | \"llm\";\n\nexport type ModelDefinition = {\n  id: string;\n  role: ModelRole;\n  kind: ModelKind;\n  dimensions?: number;\n  /** Actual deployment revision metadata; this does not pin Hugging Face requests. */\n  revision?: string;\n  schema?: string;\n  contextWindow?: number;\n  dailyCallBudget?: number;\n  supportsBatch?: boolean;\n  /**\n   * Cosine at or above which a memory counts as semantically relevant to a\n   * task on its own (without a shared word or path). Calibrated per model by\n   * `npm run eval:semantic`; cosines are not comparable across models.\n   */\n  relevanceFloor?: number;\n};\n\nexport const EMBEDDING_DIMENSIONS = 384;\n\n/**\n * Default local embedder (ONNX via transformers.js, no network after the first\n * download): ModernBERT-based, 384-d, 8k context, trained on code retrieval.\n * The revision is a pinned commit so the embedding space never drifts.\n */\nexport const LOCAL_EMBEDDING_MODEL = {\n  id: \"onnx-community/granite-embedding-small-english-r2-ONNX\",\n  role: \"primary\",\n  kind: \"embedding\",\n  dimensions: EMBEDDING_DIMENSIONS,\n  revision: \"1dc7835ba0cb9c76a3618d0bf0c427c97671b3c8\",\n  schema: \"transformers.js@4-q8-cls-l2\",\n  supportsBatch: true,\n  relevanceFloor: 0.84,\n} satisfies ModelDefinition;\n\n/**\n * Cross-encoder rerankers per `rerank.provider` (off by default: on\n * `npm run eval:semantic` none beat granite's dense order). The local default\n * was the best small one measured (13/20 first results, ~20 ms per prompt).\n */\nexport const DEFAULT_RERANK_MODELS = {\n  local: {\n    id: \"mixedbread-ai/mxbai-rerank-xsmall-v1\",\n    revision: \"b5c6e9da73abc3711f593f705371cdbe9e0fe422\",\n  },\n  huggingface: { id: \"BAAI/bge-reranker-v2-m3\" },\n  openrouter: { id: \"cohere/rerank-v3.5\" },\n} as const;\n\n/** Default hosted embedders; each must return 384-d vectors. */\nexport const DEFAULT_CLOUD_EMBEDDING_MODELS = {\n  huggingface: \"sentence-transformers/all-MiniLM-L6-v2\",\n  openrouter: \"openai/text-embedding-3-small\",\n} as const;\n\nexport const DEFAULT_EMBEDDING_MODELS = [\n  {\n    id: \"sentence-transformers/all-MiniLM-L6-v2\",\n    role: \"primary\",\n    kind: \"embedding\",\n    dimensions: EMBEDDING_DIMENSIONS,\n    dailyCallBudget: 1_000,\n    supportsBatch: true,\n    relevanceFloor: 0.42,\n  },\n  {\n    id: \"BAAI/bge-small-en-v1.5\",\n    role: \"fallback\",\n    kind: \"embedding\",\n    dimensions: EMBEDDING_DIMENSIONS,\n    dailyCallBudget: 1_000,\n    supportsBatch: true,\n    relevanceFloor: 0.69,\n  },\n] satisfies ModelDefinition[];\n\nexport const DEFAULT_LLM_MODELS = [\n  {\n    id: \"meta-llama/Llama-3.2-3B-Instruct\",\n    role: \"primary\",\n    kind: \"llm\",\n    contextWindow: 131_072,\n    dailyCallBudget: 300,\n  },\n  {\n    id: \"Qwen/Qwen2.5-7B-Instruct\",\n    role: \"fallback\",\n    kind: \"llm\",\n    contextWindow: 32_768,\n    dailyCallBudget: 300,\n  },\n  {\n    id: \"mistralai/Mistral-7B-Instruct-v0.3\",\n    role: \"last_resort\",\n    kind: \"llm\",\n    contextWindow: 32_768,\n    dailyCallBudget: 300,\n  },\n] satisfies ModelDefinition[];\n\nexport const MODEL_REGISTRY: ModelDefinition[] = [\n  LOCAL_EMBEDDING_MODEL,\n  ...DEFAULT_EMBEDDING_MODELS,\n  ...DEFAULT_LLM_MODELS,\n];\n\nexport function getModelDefinition(modelId: string): ModelDefinition | undefined {\n  return MODEL_REGISTRY.find((model) => model.id === modelId);\n}\n\nexport function getModelContextWindow(modelId: string): number | undefined {\n  return getModelDefinition(modelId)?.contextWindow;\n}\n\nexport function assertCompatibleEmbeddingModels(models: ModelDefinition[]): void {\n  const mismatchedModel = models.find((model) => model.dimensions !== EMBEDDING_DIMENSIONS);\n\n  if (mismatchedModel) {\n    throw new Error(\n      `Embedding model ${mismatchedModel.id} has ${mismatchedModel.dimensions ?? \"unknown\"} dimensions; expected ${EMBEDDING_DIMENSIONS}.`,\n    );\n  }\n}\n"],"mappings":";AAsBO,IAAM,uBAAuB;AAO7B,IAAM,wBAAwB;AAAA,EACnC,IAAI;AAAA,EACJ,MAAM;AAAA,EACN,MAAM;AAAA,EACN,YAAY;AAAA,EACZ,UAAU;AAAA,EACV,QAAQ;AAAA,EACR,eAAe;AAAA,EACf,gBAAgB;AAClB;AAOO,IAAM,wBAAwB;AAAA,EACnC,OAAO;AAAA,IACL,IAAI;AAAA,IACJ,UAAU;AAAA,EACZ;AAAA,EACA,aAAa,EAAE,IAAI,0BAA0B;AAAA,EAC7C,YAAY,EAAE,IAAI,qBAAqB;AACzC;AAGO,IAAM,iCAAiC;AAAA,EAC5C,aAAa;AAAA,EACb,YAAY;AACd;AAEO,IAAM,2BAA2B;AAAA,EACtC;AAAA,IACE,IAAI;AAAA,IACJ,MAAM;AAAA,IACN,MAAM;AAAA,IACN,YAAY;AAAA,IACZ,iBAAiB;AAAA,IACjB,eAAe;AAAA,IACf,gBAAgB;AAAA,EAClB;AAAA,EACA;AAAA,IACE,IAAI;AAAA,IACJ,MAAM;AAAA,IACN,MAAM;AAAA,IACN,YAAY;AAAA,IACZ,iBAAiB;AAAA,IACjB,eAAe;AAAA,IACf,gBAAgB;AAAA,EAClB;AACF;AAEO,IAAM,qBAAqB;AAAA,EAChC;AAAA,IACE,IAAI;AAAA,IACJ,MAAM;AAAA,IACN,MAAM;AAAA,IACN,eAAe;AAAA,IACf,iBAAiB;AAAA,EACnB;AAAA,EACA;AAAA,IACE,IAAI;AAAA,IACJ,MAAM;AAAA,IACN,MAAM;AAAA,IACN,eAAe;AAAA,IACf,iBAAiB;AAAA,EACnB;AAAA,EACA;AAAA,IACE,IAAI;AAAA,IACJ,MAAM;AAAA,IACN,MAAM;AAAA,IACN,eAAe;AAAA,IACf,iBAAiB;AAAA,EACnB;AACF;AAEO,IAAM,iBAAoC;AAAA,EAC/C;AAAA,EACA,GAAG;AAAA,EACH,GAAG;AACL;AAEO,SAAS,mBAAmB,SAA8C;AAC/E,SAAO,eAAe,KAAK,CAAC,UAAU,MAAM,OAAO,OAAO;AAC5D;AAEO,SAAS,sBAAsB,SAAqC;AACzE,SAAO,mBAAmB,OAAO,GAAG;AACtC;AAEO,SAAS,gCAAgC,QAAiC;AAC/E,QAAM,kBAAkB,OAAO,KAAK,CAAC,UAAU,MAAM,eAAe,oBAAoB;AAExF,MAAI,iBAAiB;AACnB,UAAM,IAAI;AAAA,MACR,mBAAmB,gBAAgB,EAAE,QAAQ,gBAAgB,cAAc,SAAS,yBAAyB,oBAAoB;AAAA,IACnI;AAAA,EACF;AACF;","names":[]}

package/dist/{chunk-ROQTT5ZO.js → chunk-N34XHWBA.js}

@@ -8,7 +8,7 @@ import {
   detectTestCommand,
   estimateTokens,
   findCoveringTest
-} from "./chunk-CQPCHMBS.js";
+} from "./chunk-SYWJG5NA.js";
 import {
   NON_SOURCE_PATH,
   discoverTaskPaths,
@@ -172,4 +172,4 @@ export {
   renderLeanBlock,
   runLeanHookCommand
 };
-//# sourceMappingURL=chunk-ROQTT5ZO.js.map
+//# sourceMappingURL=chunk-N34XHWBA.js.map

package/dist/{chunk-CQPCHMBS.js → chunk-SYWJG5NA.js}

@@ -1,7 +1,7 @@
 import {
   countTokens,
   generateEntityId
-} from "./chunk-4BYGTEJT.js";
+} from "./chunk-XBJQKCLU.js";
 import {
   writeAtomic
 } from "./chunk-776EQXDV.js";
@@ -316,4 +316,4 @@ export {
   findCoveringTest,
   detectTestCommand
 };
-//# sourceMappingURL=chunk-CQPCHMBS.js.map
+//# sourceMappingURL=chunk-SYWJG5NA.js.map

package/dist/{chunk-3XYQGFPB.js → chunk-U4HAJLIB.js}

@@ -1,6 +1,6 @@
 import {
   LOCAL_EMBEDDING_MODEL
-} from "./chunk-JMWJ7V6Y.js";
+} from "./chunk-JPHJBCZB.js";
 
 // src/providers/local.ts
 import { homedir } from "os";
@@ -88,4 +88,4 @@ export {
   LocalModelProvider,
   rerankLocally
 };
-//# sourceMappingURL=chunk-3XYQGFPB.js.map
+//# sourceMappingURL=chunk-U4HAJLIB.js.map

package/dist/{chunk-JDZ3VK26.js → chunk-UN57FKQX.js}

@@ -9,13 +9,13 @@ import {
   countTokens,
   generateEntityId,
   generateSessionId
-} from "./chunk-4BYGTEJT.js";
+} from "./chunk-XBJQKCLU.js";
 import {
   OpenRouterEmbeddingProvider
-} from "./chunk-KYK6RPAJ.js";
+} from "./chunk-JFDRBPYF.js";
 import {
   LocalModelProvider
-} from "./chunk-3XYQGFPB.js";
+} from "./chunk-U4HAJLIB.js";
 import {
   DEFAULT_CLOUD_EMBEDDING_MODELS,
   DEFAULT_EMBEDDING_MODELS,
@@ -23,7 +23,7 @@ import {
   EMBEDDING_DIMENSIONS,
   assertCompatibleEmbeddingModels,
   getModelDefinition
-} from "./chunk-JMWJ7V6Y.js";
+} from "./chunk-JPHJBCZB.js";
 import {
   MISSING_DATABASE_URL,
   assertNoSymlinkBetween,
@@ -549,6 +549,32 @@ function lexicalTerms(query) {
     )
   ].filter((term) => !QUERY_STOPWORDS.has(term)).slice(0, 32).map((term) => term.slice(0, 64));
 }
+var CLAIM_BOILERPLATE = /* @__PURE__ */ new Set([
+  "decided",
+  "decide",
+  "decision",
+  "made",
+  "use",
+  "uses",
+  "used",
+  "using",
+  "now",
+  "going",
+  "will"
+]);
+function termOverlap(a, b) {
+  const terms = (text2) => new Set(
+    (text2.toLowerCase().match(/[\p{L}\p{N}_]+/gu) ?? []).filter(
+      (term) => !QUERY_STOPWORDS.has(term) && !CLAIM_BOILERPLATE.has(term)
+    )
+  );
+  const left = terms(a);
+  const right = terms(b);
+  if (left.size === 0 || right.size === 0) return 0;
+  let shared = 0;
+  for (const term of left) if (right.has(term)) shared += 1;
+  return shared / (left.size + right.size - shared);
+}
 
 // src/providers/embedding-space.ts
 function embeddingSpace(response) {
@@ -3979,6 +4005,7 @@ async function checkModels(provider, skip = false) {
 export {
   isEligibleMemory,
   lexicalTerms,
+  termOverlap,
   embeddingSpaceKey,
   DefaultPolicy,
   HuggingFaceProvider,
@@ -4003,4 +4030,4 @@ export {
   inspectNamespace,
   runDoctor
 };
-//# sourceMappingURL=chunk-JDZ3VK26.js.map
+//# sourceMappingURL=chunk-UN57FKQX.js.map