@waniwani/sdk 0.1.20-beta.25 → 0.1.20-beta.26

package/dist/kb/index.d.ts

@@ -37,6 +37,14 @@ interface KnowledgeBase {
 declare function chunkMarkdown(filename: string, content: string): Omit<Chunk, "embedding">[];
 declare function generateEmbeddings(options: GenerateEmbeddingsOptions): Promise<void>;
 
-declare function loadKnowledgeBase(embeddingsPath: string): KnowledgeBase;
+/**
+ * Create a knowledge base instance from pre-loaded embeddings data.
+ *
+ * ```typescript
+ * import embeddings from "./embeddings.json";
+ * const kb = loadKnowledgeBase(embeddings);
+ * ```
+ */
+declare function loadKnowledgeBase(source: EmbeddingsFile): KnowledgeBase;
 
 export { type Chunk, type EmbeddingsFile, type GenerateEmbeddingsOptions, type KnowledgeBase, type SearchResult, chunkMarkdown, generateEmbeddings, loadKnowledgeBase };

package/dist/kb/index.js

@@ -1,11 +1,11 @@
-import{readdir as f,readFile as b,writeFile as k}from"fs/promises";import{join as $}from"path";import{embedMany as E}from"ai";var w="openai/text-embedding-3-small",y=512;function p(o,t){let l=t.split(`
-`),i=[],r=t.match(/^# (.+)$/m)?.[1]??o.replace(".md",""),n=r,s=[],d=0;for(let e of l){if(e.match(/^# /))continue;let c=e.match(/^## (.+)$/);if(c){if(s.length>0){let g=s.join(`
-`).trim();g&&(i.push({id:`${o.replace(".md","")}#${d}`,source:o,heading:n,content:`${r}: ${n}
+import{readdir as b,readFile as f,writeFile as k}from"fs/promises";import{join as $}from"path";import{embedMany as w}from"ai";var E="openai/text-embedding-3-small",y=512;function p(o,d){let a=d.split(`
+`),i=[],s=d.match(/^# (.+)$/m)?.[1]??o.replace(".md",""),n=s,r=[],c=0;for(let e of a){if(e.match(/^# /))continue;let m=e.match(/^## (.+)$/);if(m){if(r.length>0){let g=r.join(`
+`).trim();g&&(i.push({id:`${o.replace(".md","")}#${c}`,source:o,heading:n,content:`${s}: ${n}
 
-${g}`}),d++)}n=c[1],s=[]}else s.push(e)}let m=s.join(`
-`).trim();return m&&i.push({id:`${o.replace(".md","")}#${d}`,source:o,heading:n,content:`${r}: ${n}
+${g}`}),c++)}n=m[1],r=[]}else r.push(e)}let t=r.join(`
+`).trim();return t&&i.push({id:`${o.replace(".md","")}#${c}`,source:o,heading:n,content:`${s}: ${n}
 
-${m}`}),i}async function F(o){let{knowledgeDir:t,outputPath:l,model:i=w,dimensions:a=y}=o,r=(await f(t)).filter(e=>e.endsWith(".md"));console.log(`Found ${r.length} knowledge files`);let n=[];for(let e of r){let c=await b($(t,e),"utf-8"),g=p(e,c);n.push(...g),console.log(`  ${e}: ${g.length} chunk(s)`)}console.log(`
-Total chunks: ${n.length}`),console.log(`Generating embeddings with ${i}...`);let{embeddings:s}=await E({model:i,values:n.map(e=>e.content),providerOptions:{openai:{dimensions:a}}}),d=n.map((e,c)=>({...e,embedding:s[c]})),m={model:i,dimensions:a,generatedAt:new Date().toISOString(),chunks:d};await k(l,JSON.stringify(m)),console.log(`
-Written ${l} (${d.length} chunks)`)}import{readFileSync as O}from"fs";import{cosineSimilarity as S,embed as x}from"ai";function C(o){let t=null;function l(){return t||(t=JSON.parse(O(o,"utf-8")),t)}return{get chunkCount(){return l().chunks.length},async search(i,a=5){let{model:r,dimensions:n,chunks:s}=l(),{embedding:d}=await x({model:r,value:i,...n&&{providerOptions:{openai:{dimensions:n}}}});return s.map(e=>({source:e.source,heading:e.heading,content:e.content,score:S(d,e.embedding)})).sort((e,c)=>c.score-e.score).slice(0,a)}}}export{p as chunkMarkdown,F as generateEmbeddings,C as loadKnowledgeBase};
+${t}`}),i}async function O(o){let{knowledgeDir:d,outputPath:a,model:i=E,dimensions:l=y}=o,s=(await b(d)).filter(e=>e.endsWith(".md"));console.log(`Found ${s.length} knowledge files`);let n=[];for(let e of s){let m=await f($(d,e),"utf-8"),g=p(e,m);n.push(...g),console.log(`  ${e}: ${g.length} chunk(s)`)}console.log(`
+Total chunks: ${n.length}`),console.log(`Generating embeddings with ${i}...`);let{embeddings:r}=await w({model:i,values:n.map(e=>e.content),providerOptions:{openai:{dimensions:l}}}),c=n.map((e,m)=>({...e,embedding:r[m]})),t={model:i,dimensions:l,generatedAt:new Date().toISOString(),chunks:c};await k(a,JSON.stringify(t)),console.log(`
+Written ${a} (${c.length} chunks)`)}import{cosineSimilarity as F,embed as x}from"ai";function C(o){function d(){return o}return{get chunkCount(){return d().chunks.length},async search(a,i=5){let{model:l,dimensions:s,chunks:n}=d(),{embedding:r}=await x({model:l,value:a,...s&&{providerOptions:{openai:{dimensions:s}}}});return n.map(t=>({source:t.source,heading:t.heading,content:t.content,score:F(r,t.embedding)})).sort((t,e)=>e.score-t.score).slice(0,i)}}}export{p as chunkMarkdown,O as generateEmbeddings,C as loadKnowledgeBase};
 //# sourceMappingURL=index.js.map

package/dist/kb/index.js.map

@@ -1 +1 @@
-{"version":3,"sources":["../../src/kb/embed.ts","../../src/kb/search.ts"],"sourcesContent":["import { readdir, readFile, writeFile } from \"node:fs/promises\";\nimport { join } from \"node:path\";\nimport { embedMany } from \"ai\";\nimport type { Chunk, EmbeddingsFile, GenerateEmbeddingsOptions } from \"./types\";\n\nconst DEFAULT_MODEL = \"openai/text-embedding-3-small\";\nconst DEFAULT_DIMENSIONS = 512;\n\nexport function chunkMarkdown(\n\tfilename: string,\n\tcontent: string,\n): Omit<Chunk, \"embedding\">[] {\n\tconst lines = content.split(\"\\n\");\n\tconst chunks: Omit<Chunk, \"embedding\">[] = [];\n\n\tconst h1Match = content.match(/^# (.+)$/m);\n\tconst title = h1Match?.[1] ?? filename.replace(\".md\", \"\");\n\n\tlet currentHeading = title;\n\tlet currentContent: string[] = [];\n\tlet chunkIndex = 0;\n\n\tfor (const line of lines) {\n\t\tif (line.match(/^# /)) continue;\n\n\t\tconst h2Match = line.match(/^## (.+)$/);\n\t\tif (h2Match) {\n\t\t\tif (currentContent.length > 0) {\n\t\t\t\tconst text = currentContent.join(\"\\n\").trim();\n\t\t\t\tif (text) {\n\t\t\t\t\tchunks.push({\n\t\t\t\t\t\tid: `${filename.replace(\".md\", \"\")}#${chunkIndex}`,\n\t\t\t\t\t\tsource: filename,\n\t\t\t\t\t\theading: currentHeading,\n\t\t\t\t\t\tcontent: `${title}: ${currentHeading}\\n\\n${text}`,\n\t\t\t\t\t});\n\t\t\t\t\tchunkIndex++;\n\t\t\t\t}\n\t\t\t}\n\t\t\tcurrentHeading = h2Match[1];\n\t\t\tcurrentContent = [];\n\t\t} else {\n\t\t\tcurrentContent.push(line);\n\t\t}\n\t}\n\n\tconst text = currentContent.join(\"\\n\").trim();\n\tif (text) {\n\t\tchunks.push({\n\t\t\tid: `${filename.replace(\".md\", \"\")}#${chunkIndex}`,\n\t\t\tsource: filename,\n\t\t\theading: currentHeading,\n\t\t\tcontent: `${title}: ${currentHeading}\\n\\n${text}`,\n\t\t});\n\t}\n\n\treturn chunks;\n}\n\nexport async function generateEmbeddings(\n\toptions: GenerateEmbeddingsOptions,\n): Promise<void> {\n\tconst {\n\t\tknowledgeDir,\n\t\toutputPath,\n\t\tmodel = DEFAULT_MODEL,\n\t\tdimensions = DEFAULT_DIMENSIONS,\n\t} = options;\n\n\tconst files = (await readdir(knowledgeDir)).filter((f) => f.endsWith(\".md\"));\n\tconsole.log(`Found ${files.length} knowledge files`);\n\n\tconst allChunks: Omit<Chunk, \"embedding\">[] = [];\n\tfor (const file of files) {\n\t\tconst content = await readFile(join(knowledgeDir, file), \"utf-8\");\n\t\tconst chunks = chunkMarkdown(file, content);\n\t\tallChunks.push(...chunks);\n\t\tconsole.log(`  ${file}: ${chunks.length} chunk(s)`);\n\t}\n\n\tconsole.log(`\\nTotal chunks: ${allChunks.length}`);\n\tconsole.log(`Generating embeddings with ${model}...`);\n\n\tconst { embeddings } = await embedMany({\n\t\tmodel,\n\t\tvalues: allChunks.map((c) => c.content),\n\t\tproviderOptions: { openai: { dimensions } },\n\t});\n\n\tconst chunks: Chunk[] = allChunks.map((chunk, i) => ({\n\t\t...chunk,\n\t\tembedding: embeddings[i],\n\t}));\n\n\tconst output: EmbeddingsFile = {\n\t\tmodel,\n\t\tdimensions,\n\t\tgeneratedAt: new Date().toISOString(),\n\t\tchunks,\n\t};\n\n\tawait writeFile(outputPath, JSON.stringify(output));\n\tconsole.log(`\\nWritten ${outputPath} (${chunks.length} chunks)`);\n}\n","import { readFileSync } from \"node:fs\";\nimport { cosineSimilarity, embed } from \"ai\";\nimport type { EmbeddingsFile, KnowledgeBase, SearchResult } from \"./types\";\n\nexport function loadKnowledgeBase(embeddingsPath: string): KnowledgeBase {\n\tlet cached: EmbeddingsFile | null = null;\n\n\tfunction getEmbeddings(): EmbeddingsFile {\n\t\tif (cached) return cached;\n\t\tcached = JSON.parse(\n\t\t\treadFileSync(embeddingsPath, \"utf-8\"),\n\t\t) as EmbeddingsFile;\n\t\treturn cached;\n\t}\n\n\treturn {\n\t\tget chunkCount() {\n\t\t\treturn getEmbeddings().chunks.length;\n\t\t},\n\n\t\tasync search(query: string, topK = 5): Promise<SearchResult[]> {\n\t\t\tconst { model, dimensions, chunks } = getEmbeddings();\n\n\t\t\tconst { embedding: queryEmbedding } = await embed({\n\t\t\t\tmodel,\n\t\t\t\tvalue: query,\n\t\t\t\t...(dimensions && {\n\t\t\t\t\tproviderOptions: { openai: { dimensions } },\n\t\t\t\t}),\n\t\t\t});\n\n\t\t\tconst scored = chunks.map((chunk) => ({\n\t\t\t\tsource: chunk.source,\n\t\t\t\theading: chunk.heading,\n\t\t\t\tcontent: chunk.content,\n\t\t\t\tscore: cosineSimilarity(queryEmbedding, chunk.embedding),\n\t\t\t}));\n\n\t\t\treturn scored.sort((a, b) => b.score - a.score).slice(0, topK);\n\t\t},\n\t};\n}\n"],"mappings":"AAAA,OAAS,WAAAA,EAAS,YAAAC,EAAU,aAAAC,MAAiB,cAC7C,OAAS,QAAAC,MAAY,OACrB,OAAS,aAAAC,MAAiB,KAG1B,IAAMC,EAAgB,gCAChBC,EAAqB,IAEpB,SAASC,EACfC,EACAC,EAC6B,CAC7B,IAAMC,EAAQD,EAAQ,MAAM;AAAA,CAAI,EAC1BE,EAAqC,CAAC,EAGtCC,EADUH,EAAQ,MAAM,WAAW,IACjB,CAAC,GAAKD,EAAS,QAAQ,MAAO,EAAE,EAEpDK,EAAiBD,EACjBE,EAA2B,CAAC,EAC5BC,EAAa,EAEjB,QAAWC,KAAQN,EAAO,CACzB,GAAIM,EAAK,MAAM,KAAK,EAAG,SAEvB,IAAMC,EAAUD,EAAK,MAAM,WAAW,EACtC,GAAIC,EAAS,CACZ,GAAIH,EAAe,OAAS,EAAG,CAC9B,IAAMI,EAAOJ,EAAe,KAAK;AAAA,CAAI,EAAE,KAAK,EACxCI,IACHP,EAAO,KAAK,CACX,GAAI,GAAGH,EAAS,QAAQ,MAAO,EAAE,CAAC,IAAIO,CAAU,GAChD,OAAQP,EACR,QAASK,EACT,QAAS,GAAGD,CAAK,KAAKC,CAAc;AAAA;AAAA,EAAOK,CAAI,EAChD,CAAC,EACDH,IAEF,CACAF,EAAiBI,EAAQ,CAAC,EAC1BH,EAAiB,CAAC,CACnB,MACCA,EAAe,KAAKE,CAAI,CAE1B,CAEA,IAAME,EAAOJ,EAAe,KAAK;AAAA,CAAI,EAAE,KAAK,EAC5C,OAAII,GACHP,EAAO,KAAK,CACX,GAAI,GAAGH,EAAS,QAAQ,MAAO,EAAE,CAAC,IAAIO,CAAU,GAChD,OAAQP,EACR,QAASK,EACT,QAAS,GAAGD,CAAK,KAAKC,CAAc;AAAA;AAAA,EAAOK,CAAI,EAChD,CAAC,EAGKP,CACR,CAEA,eAAsBQ,EACrBC,EACgB,CAChB,GAAM,CACL,aAAAC,EACA,WAAAC,EACA,MAAAC,EAAQlB,EACR,WAAAmB,EAAalB,CACd,EAAIc,EAEEK,GAAS,MAAMzB,EAAQqB,CAAY,GAAG,OAAQK,GAAMA,EAAE,SAAS,KAAK,CAAC,EAC3E,QAAQ,IAAI,SAASD,EAAM,MAAM,kBAAkB,EAEnD,IAAME,EAAwC,CAAC,EAC/C,QAAWC,KAAQH,EAAO,CACzB,IAAMhB,EAAU,MAAMR,EAASE,EAAKkB,EAAcO,CAAI,EAAG,OAAO,EAC1DjB,EAASJ,EAAcqB,EAAMnB,CAAO,EAC1CkB,EAAU,KAAK,GAAGhB,CAAM,EACxB,QAAQ,IAAI,KAAKiB,CAAI,KAAKjB,EAAO,MAAM,WAAW,CACnD,CAEA,QAAQ,IAAI;AAAA,gBAAmBgB,EAAU,MAAM,EAAE,EACjD,QAAQ,IAAI,8BAA8BJ,CAAK,KAAK,EAEpD,GAAM,CAAE,WAAAM,CAAW,EAAI,MAAMzB,EAAU,CACtC,MAAAmB,EACA,OAAQI,EAAU,IAAKG,GAAMA,EAAE,OAAO,EACtC,gBAAiB,CAAE,OAAQ,CAAE,WAAAN,CAAW,CAAE,CAC3C,CAAC,EAEKb,EAAkBgB,EAAU,IAAI,CAACI,EAAOC,KAAO,CACpD,GAAGD,EACH,UAAWF,EAAWG,CAAC,CACxB,EAAE,EAEIC,EAAyB,CAC9B,MAAAV,EACA,WAAAC,EACA,YAAa,IAAI,KAAK,EAAE,YAAY,EACpC,OAAAb,CACD,EAEA,MAAMT,EAAUoB,EAAY,KAAK,UAAUW,CAAM,CAAC,EAClD,QAAQ,IAAI;AAAA,UAAaX,CAAU,KAAKX,EAAO,MAAM,UAAU,CAChE,CCvGA,OAAS,gBAAAuB,MAAoB,KAC7B,OAAS,oBAAAC,EAAkB,SAAAC,MAAa,KAGjC,SAASC,EAAkBC,EAAuC,CACxE,IAAIC,EAAgC,KAEpC,SAASC,GAAgC,CACxC,OAAID,IACJA,EAAS,KAAK,MACbL,EAAaI,EAAgB,OAAO,CACrC,EACOC,EACR,CAEA,MAAO,CACN,IAAI,YAAa,CAChB,OAAOC,EAAc,EAAE,OAAO,MAC/B,EAEA,MAAM,OAAOC,EAAeC,EAAO,EAA4B,CAC9D,GAAM,CAAE,MAAAC,EAAO,WAAAC,EAAY,OAAAC,CAAO,EAAIL,EAAc,EAE9C,CAAE,UAAWM,CAAe,EAAI,MAAMV,EAAM,CACjD,MAAAO,EACA,MAAOF,EACP,GAAIG,GAAc,CACjB,gBAAiB,CAAE,OAAQ,CAAE,WAAAA,CAAW,CAAE,CAC3C,CACD,CAAC,EASD,OAPeC,EAAO,IAAKE,IAAW,CACrC,OAAQA,EAAM,OACd,QAASA,EAAM,QACf,QAASA,EAAM,QACf,MAAOZ,EAAiBW,EAAgBC,EAAM,SAAS,CACxD,EAAE,EAEY,KAAK,CAACC,EAAGC,IAAMA,EAAE,MAAQD,EAAE,KAAK,EAAE,MAAM,EAAGN,CAAI,CAC9D,CACD,CACD","names":["readdir","readFile","writeFile","join","embedMany","DEFAULT_MODEL","DEFAULT_DIMENSIONS","chunkMarkdown","filename","content","lines","chunks","title","currentHeading","currentContent","chunkIndex","line","h2Match","text","generateEmbeddings","options","knowledgeDir","outputPath","model","dimensions","files","f","allChunks","file","embeddings","c","chunk","i","output","readFileSync","cosineSimilarity","embed","loadKnowledgeBase","embeddingsPath","cached","getEmbeddings","query","topK","model","dimensions","chunks","queryEmbedding","chunk","a","b"]}
+{"version":3,"sources":["../../src/kb/embed.ts","../../src/kb/search.ts"],"sourcesContent":["import { readdir, readFile, writeFile } from \"node:fs/promises\";\nimport { join } from \"node:path\";\nimport { embedMany } from \"ai\";\nimport type { Chunk, EmbeddingsFile, GenerateEmbeddingsOptions } from \"./types\";\n\nconst DEFAULT_MODEL = \"openai/text-embedding-3-small\";\nconst DEFAULT_DIMENSIONS = 512;\n\nexport function chunkMarkdown(\n\tfilename: string,\n\tcontent: string,\n): Omit<Chunk, \"embedding\">[] {\n\tconst lines = content.split(\"\\n\");\n\tconst chunks: Omit<Chunk, \"embedding\">[] = [];\n\n\tconst h1Match = content.match(/^# (.+)$/m);\n\tconst title = h1Match?.[1] ?? filename.replace(\".md\", \"\");\n\n\tlet currentHeading = title;\n\tlet currentContent: string[] = [];\n\tlet chunkIndex = 0;\n\n\tfor (const line of lines) {\n\t\tif (line.match(/^# /)) continue;\n\n\t\tconst h2Match = line.match(/^## (.+)$/);\n\t\tif (h2Match) {\n\t\t\tif (currentContent.length > 0) {\n\t\t\t\tconst text = currentContent.join(\"\\n\").trim();\n\t\t\t\tif (text) {\n\t\t\t\t\tchunks.push({\n\t\t\t\t\t\tid: `${filename.replace(\".md\", \"\")}#${chunkIndex}`,\n\t\t\t\t\t\tsource: filename,\n\t\t\t\t\t\theading: currentHeading,\n\t\t\t\t\t\tcontent: `${title}: ${currentHeading}\\n\\n${text}`,\n\t\t\t\t\t});\n\t\t\t\t\tchunkIndex++;\n\t\t\t\t}\n\t\t\t}\n\t\t\tcurrentHeading = h2Match[1];\n\t\t\tcurrentContent = [];\n\t\t} else {\n\t\t\tcurrentContent.push(line);\n\t\t}\n\t}\n\n\tconst text = currentContent.join(\"\\n\").trim();\n\tif (text) {\n\t\tchunks.push({\n\t\t\tid: `${filename.replace(\".md\", \"\")}#${chunkIndex}`,\n\t\t\tsource: filename,\n\t\t\theading: currentHeading,\n\t\t\tcontent: `${title}: ${currentHeading}\\n\\n${text}`,\n\t\t});\n\t}\n\n\treturn chunks;\n}\n\nexport async function generateEmbeddings(\n\toptions: GenerateEmbeddingsOptions,\n): Promise<void> {\n\tconst {\n\t\tknowledgeDir,\n\t\toutputPath,\n\t\tmodel = DEFAULT_MODEL,\n\t\tdimensions = DEFAULT_DIMENSIONS,\n\t} = options;\n\n\tconst files = (await readdir(knowledgeDir)).filter((f) => f.endsWith(\".md\"));\n\tconsole.log(`Found ${files.length} knowledge files`);\n\n\tconst allChunks: Omit<Chunk, \"embedding\">[] = [];\n\tfor (const file of files) {\n\t\tconst content = await readFile(join(knowledgeDir, file), \"utf-8\");\n\t\tconst chunks = chunkMarkdown(file, content);\n\t\tallChunks.push(...chunks);\n\t\tconsole.log(`  ${file}: ${chunks.length} chunk(s)`);\n\t}\n\n\tconsole.log(`\\nTotal chunks: ${allChunks.length}`);\n\tconsole.log(`Generating embeddings with ${model}...`);\n\n\tconst { embeddings } = await embedMany({\n\t\tmodel,\n\t\tvalues: allChunks.map((c) => c.content),\n\t\tproviderOptions: { openai: { dimensions } },\n\t});\n\n\tconst chunks: Chunk[] = allChunks.map((chunk, i) => ({\n\t\t...chunk,\n\t\tembedding: embeddings[i],\n\t}));\n\n\tconst output: EmbeddingsFile = {\n\t\tmodel,\n\t\tdimensions,\n\t\tgeneratedAt: new Date().toISOString(),\n\t\tchunks,\n\t};\n\n\tawait writeFile(outputPath, JSON.stringify(output));\n\tconsole.log(`\\nWritten ${outputPath} (${chunks.length} chunks)`);\n}\n","import { cosineSimilarity, embed } from \"ai\";\nimport type { EmbeddingsFile, KnowledgeBase, SearchResult } from \"./types\";\n\n/**\n * Create a knowledge base instance from pre-loaded embeddings data.\n *\n * ```typescript\n * import embeddings from \"./embeddings.json\";\n * const kb = loadKnowledgeBase(embeddings);\n * ```\n */\nexport function loadKnowledgeBase(source: EmbeddingsFile): KnowledgeBase {\n\tfunction getEmbeddings(): EmbeddingsFile {\n\t\treturn source;\n\t}\n\n\treturn {\n\t\tget chunkCount() {\n\t\t\treturn getEmbeddings().chunks.length;\n\t\t},\n\n\t\tasync search(query: string, topK = 5): Promise<SearchResult[]> {\n\t\t\tconst { model, dimensions, chunks } = getEmbeddings();\n\n\t\t\tconst { embedding: queryEmbedding } = await embed({\n\t\t\t\tmodel,\n\t\t\t\tvalue: query,\n\t\t\t\t...(dimensions && {\n\t\t\t\t\tproviderOptions: { openai: { dimensions } },\n\t\t\t\t}),\n\t\t\t});\n\n\t\t\tconst scored = chunks.map((chunk) => ({\n\t\t\t\tsource: chunk.source,\n\t\t\t\theading: chunk.heading,\n\t\t\t\tcontent: chunk.content,\n\t\t\t\tscore: cosineSimilarity(queryEmbedding, chunk.embedding),\n\t\t\t}));\n\n\t\t\treturn scored.sort((a, b) => b.score - a.score).slice(0, topK);\n\t\t},\n\t};\n}\n"],"mappings":"AAAA,OAAS,WAAAA,EAAS,YAAAC,EAAU,aAAAC,MAAiB,cAC7C,OAAS,QAAAC,MAAY,OACrB,OAAS,aAAAC,MAAiB,KAG1B,IAAMC,EAAgB,gCAChBC,EAAqB,IAEpB,SAASC,EACfC,EACAC,EAC6B,CAC7B,IAAMC,EAAQD,EAAQ,MAAM;AAAA,CAAI,EAC1BE,EAAqC,CAAC,EAGtCC,EADUH,EAAQ,MAAM,WAAW,IACjB,CAAC,GAAKD,EAAS,QAAQ,MAAO,EAAE,EAEpDK,EAAiBD,EACjBE,EAA2B,CAAC,EAC5BC,EAAa,EAEjB,QAAWC,KAAQN,EAAO,CACzB,GAAIM,EAAK,MAAM,KAAK,EAAG,SAEvB,IAAMC,EAAUD,EAAK,MAAM,WAAW,EACtC,GAAIC,EAAS,CACZ,GAAIH,EAAe,OAAS,EAAG,CAC9B,IAAMI,EAAOJ,EAAe,KAAK;AAAA,CAAI,EAAE,KAAK,EACxCI,IACHP,EAAO,KAAK,CACX,GAAI,GAAGH,EAAS,QAAQ,MAAO,EAAE,CAAC,IAAIO,CAAU,GAChD,OAAQP,EACR,QAASK,EACT,QAAS,GAAGD,CAAK,KAAKC,CAAc;AAAA;AAAA,EAAOK,CAAI,EAChD,CAAC,EACDH,IAEF,CACAF,EAAiBI,EAAQ,CAAC,EAC1BH,EAAiB,CAAC,CACnB,MACCA,EAAe,KAAKE,CAAI,CAE1B,CAEA,IAAME,EAAOJ,EAAe,KAAK;AAAA,CAAI,EAAE,KAAK,EAC5C,OAAII,GACHP,EAAO,KAAK,CACX,GAAI,GAAGH,EAAS,QAAQ,MAAO,EAAE,CAAC,IAAIO,CAAU,GAChD,OAAQP,EACR,QAASK,EACT,QAAS,GAAGD,CAAK,KAAKC,CAAc;AAAA;AAAA,EAAOK,CAAI,EAChD,CAAC,EAGKP,CACR,CAEA,eAAsBQ,EACrBC,EACgB,CAChB,GAAM,CACL,aAAAC,EACA,WAAAC,EACA,MAAAC,EAAQlB,EACR,WAAAmB,EAAalB,CACd,EAAIc,EAEEK,GAAS,MAAMzB,EAAQqB,CAAY,GAAG,OAAQK,GAAMA,EAAE,SAAS,KAAK,CAAC,EAC3E,QAAQ,IAAI,SAASD,EAAM,MAAM,kBAAkB,EAEnD,IAAME,EAAwC,CAAC,EAC/C,QAAWC,KAAQH,EAAO,CACzB,IAAMhB,EAAU,MAAMR,EAASE,EAAKkB,EAAcO,CAAI,EAAG,OAAO,EAC1DjB,EAASJ,EAAcqB,EAAMnB,CAAO,EAC1CkB,EAAU,KAAK,GAAGhB,CAAM,EACxB,QAAQ,IAAI,KAAKiB,CAAI,KAAKjB,EAAO,MAAM,WAAW,CACnD,CAEA,QAAQ,IAAI;AAAA,gBAAmBgB,EAAU,MAAM,EAAE,EACjD,QAAQ,IAAI,8BAA8BJ,CAAK,KAAK,EAEpD,GAAM,CAAE,WAAAM,CAAW,EAAI,MAAMzB,EAAU,CACtC,MAAAmB,EACA,OAAQI,EAAU,IAAKG,GAAMA,EAAE,OAAO,EACtC,gBAAiB,CAAE,OAAQ,CAAE,WAAAN,CAAW,CAAE,CAC3C,CAAC,EAEKb,EAAkBgB,EAAU,IAAI,CAACI,EAAOC,KAAO,CACpD,GAAGD,EACH,UAAWF,EAAWG,CAAC,CACxB,EAAE,EAEIC,EAAyB,CAC9B,MAAAV,EACA,WAAAC,EACA,YAAa,IAAI,KAAK,EAAE,YAAY,EACpC,OAAAb,CACD,EAEA,MAAMT,EAAUoB,EAAY,KAAK,UAAUW,CAAM,CAAC,EAClD,QAAQ,IAAI;AAAA,UAAaX,CAAU,KAAKX,EAAO,MAAM,UAAU,CAChE,CCvGA,OAAS,oBAAAuB,EAAkB,SAAAC,MAAa,KAWjC,SAASC,EAAkBC,EAAuC,CACxE,SAASC,GAAgC,CACxC,OAAOD,CACR,CAEA,MAAO,CACN,IAAI,YAAa,CAChB,OAAOC,EAAc,EAAE,OAAO,MAC/B,EAEA,MAAM,OAAOC,EAAeC,EAAO,EAA4B,CAC9D,GAAM,CAAE,MAAAC,EAAO,WAAAC,EAAY,OAAAC,CAAO,EAAIL,EAAc,EAE9C,CAAE,UAAWM,CAAe,EAAI,MAAMT,EAAM,CACjD,MAAAM,EACA,MAAOF,EACP,GAAIG,GAAc,CACjB,gBAAiB,CAAE,OAAQ,CAAE,WAAAA,CAAW,CAAE,CAC3C,CACD,CAAC,EASD,OAPeC,EAAO,IAAKE,IAAW,CACrC,OAAQA,EAAM,OACd,QAASA,EAAM,QACf,QAASA,EAAM,QACf,MAAOX,EAAiBU,EAAgBC,EAAM,SAAS,CACxD,EAAE,EAEY,KAAK,CAACC,EAAGC,IAAMA,EAAE,MAAQD,EAAE,KAAK,EAAE,MAAM,EAAGN,CAAI,CAC9D,CACD,CACD","names":["readdir","readFile","writeFile","join","embedMany","DEFAULT_MODEL","DEFAULT_DIMENSIONS","chunkMarkdown","filename","content","lines","chunks","title","currentHeading","currentContent","chunkIndex","line","h2Match","text","generateEmbeddings","options","knowledgeDir","outputPath","model","dimensions","files","f","allChunks","file","embeddings","c","chunk","i","output","cosineSimilarity","embed","loadKnowledgeBase","source","getEmbeddings","query","topK","model","dimensions","chunks","queryEmbedding","chunk","a","b"]}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@waniwani/sdk",
-  "version": "0.1.20-beta.25",
+  "version": "0.1.20-beta.26",
   "description": "WaniWani SDK - MCP event tracking, widget framework, and tools",
   "type": "module",
   "exports": {