@lunora/ai 1.0.0-alpha.64 → 1.0.0-alpha.66

package/dist/index.d.mts

@@ -63,9 +63,14 @@ interface ModelUsage {
  * An indicative price table, keyed by model id.
  *
  * Deliberately small: a table that tries to cover every model is a table that
- * is wrong about most of them. It holds the Workers AI models Lunora's own
- * defaults reference plus the common OpenAI embedding models, and everything
- * else returns `undefined` rather than a guess.
+ * is wrong about most of them. It holds the models Lunora's own defaults and
+ * documented examples reference — text generation and embeddings, on Workers AI
+ * and OpenAI — and everything else returns `undefined` rather than a guess.
+ *
+ * Generation models carry an `output` price; embedding models do not (they have
+ * no completion). A generation model missing from here is why a chat span would
+ * carry no cost at all off an AI Gateway, so the ids the docs teach are the ones
+ * that have to be in the table.
  */
 declare const DEFAULT_MODEL_PRICES: Readonly<Record<string, ModelPrice>>;
 /**

package/dist/index.d.ts

@@ -63,9 +63,14 @@ interface ModelUsage {
  * An indicative price table, keyed by model id.
  *
  * Deliberately small: a table that tries to cover every model is a table that
- * is wrong about most of them. It holds the Workers AI models Lunora's own
- * defaults reference plus the common OpenAI embedding models, and everything
- * else returns `undefined` rather than a guess.
+ * is wrong about most of them. It holds the models Lunora's own defaults and
+ * documented examples reference — text generation and embeddings, on Workers AI
+ * and OpenAI — and everything else returns `undefined` rather than a guess.
+ *
+ * Generation models carry an `output` price; embedding models do not (they have
+ * no completion). A generation model missing from here is why a chat span would
+ * carry no cost at all off an AI Gateway, so the ids the docs teach are the ones
+ * that have to be in the table.
  */
 declare const DEFAULT_MODEL_PRICES: Readonly<Record<string, ModelPrice>>;
 /**

package/dist/index.mjs

@@ -1 +1 @@
-import{default as o}from"./packem_shared/createAi-DIJS_f7H.mjs";import{AI_GATEWAY_ACCOUNT_ID_ENV as r,AI_GATEWAY_ID_ENV as A,AI_GATEWAY_TOKEN_ENV as _,buildAiGatewayMetadataFields as l,resolveAiGateway as m}from"./packem_shared/AI_GATEWAY_ACCOUNT_ID_ENV-9ptmPAYP.mjs";import{DEFAULT_MODEL_PRICES as E,estimateModelCost as T,lookupModelPrice as d}from"./packem_shared/DEFAULT_MODEL_PRICES-DZ3dYVUn.mjs";import{embed as x,embedMany as I,generateObject as c,generateText as f,hasToolCall as p,jsonSchema as b,streamObject as C,streamText as G,tool as M}from"ai";import{createWorkersAI as O}from"workers-ai-provider";export{r as AI_GATEWAY_ACCOUNT_ID_ENV,A as AI_GATEWAY_ID_ENV,_ as AI_GATEWAY_TOKEN_ENV,E as DEFAULT_MODEL_PRICES,l as buildAiGatewayMetadataFields,o as createAi,O as createWorkersAI,x as embed,I as embedMany,T as estimateModelCost,c as generateObject,f as generateText,p as hasToolCall,b as jsonSchema,d as lookupModelPrice,m as resolveAiGateway,C as streamObject,G as streamText,M as tool};
+import{default as o}from"./packem_shared/createAi-DIJS_f7H.mjs";import{AI_GATEWAY_ACCOUNT_ID_ENV as r,AI_GATEWAY_ID_ENV as A,AI_GATEWAY_TOKEN_ENV as _,buildAiGatewayMetadataFields as l,resolveAiGateway as m}from"./packem_shared/AI_GATEWAY_ACCOUNT_ID_ENV-9ptmPAYP.mjs";import{DEFAULT_MODEL_PRICES as E,estimateModelCost as T,lookupModelPrice as d}from"./packem_shared/DEFAULT_MODEL_PRICES-Q8uxdiuV.mjs";import{embed as x,embedMany as I,generateObject as c,generateText as f,hasToolCall as p,jsonSchema as b,streamObject as C,streamText as G,tool as M}from"ai";import{createWorkersAI as O}from"workers-ai-provider";export{r as AI_GATEWAY_ACCOUNT_ID_ENV,A as AI_GATEWAY_ID_ENV,_ as AI_GATEWAY_TOKEN_ENV,E as DEFAULT_MODEL_PRICES,l as buildAiGatewayMetadataFields,o as createAi,O as createWorkersAI,x as embed,I as embedMany,T as estimateModelCost,c as generateObject,f as generateText,p as hasToolCall,b as jsonSchema,d as lookupModelPrice,m as resolveAiGateway,C as streamObject,G as streamText,M as tool};

package/dist/packem_shared/DEFAULT_MODEL_PRICES-Q8uxdiuV.mjs

@@ -0,0 +1 @@
+const p=/^(.*)-\d{4}-\d{2}-\d{2}$/u,a={"@cf/baai/bge-base-en-v1.5":{input:.067},"@cf/baai/bge-large-en-v1.5":{input:.204},"@cf/baai/bge-m3":{input:.012},"@cf/baai/bge-small-en-v1.5":{input:.02},"@cf/meta/llama-3.1-8b-instruct":{input:.28,output:.83},"@cf/meta/llama-3.3-70b-instruct-fp8-fast":{input:.29,output:2.25},"text-embedding-3-large":{input:.13},"text-embedding-3-small":{input:.02},"gpt-4o":{input:2.5,output:10},"gpt-4o-mini":{input:.15,output:.6},"gpt-5":{input:1.25,output:10}},r=n=>{const t=n.trim(),i=t.lastIndexOf("/");return(i!==-1&&!t.startsWith("@")?[t,t.slice(i+1)]:[t]).flatMap(e=>{const o=p.exec(e)?.[1];return o===void 0?[e]:[e,o]})},c=(n,t=a)=>{for(const i of r(n))if(Object.hasOwn(t,i))return t[i]},b=(n,t,i)=>{if(n===void 0||n.length===0)return;const u=c(n,i);if(u===void 0)return;const e=Number.isFinite(t.inputTokens)?Math.max(0,t.inputTokens):0,o=Number.isFinite(t.outputTokens)?Math.max(0,t.outputTokens):0;if(e<=0&&o<=0)return;const s=(e*u.input+o*(u.output??0))/1e6;return Number.isFinite(s)?s:void 0};export{a as DEFAULT_MODEL_PRICES,b as estimateModelCost,c as lookupModelPrice};

package/dist/packem_shared/{defineRag-Dzh-MbCu.mjs → defineRag-CMM17YeT.mjs}

@@ -1,4 +1,4 @@
-import{LunoraError as g,isLunoraError as ye}from"@lunora/errors";import{tool as xe,jsonSchema as Ie,embedMany as Se,embed as _e}from"ai";import{estimateModelCost as ke}from"./DEFAULT_MODEL_PRICES-DZ3dYVUn.mjs";import Te from"./fixedWindowChunks-C461ahRE.mjs";import{c as Ae,I as Ne}from"./concurrent-C6nqBv41.mjs";import{contentHash as Me}from"./contentHash-BIn6ECP8.mjs";import ee from"./hybridRank-DejmVw2I.mjs";import{VECTORIZE_CAPABILITIES as te,vectorizeStore as De}from"./VECTORIZE_CAPABILITIES-CUQDoxis.mjs";const Re=1e3,Ce=200,$e=5,Be=4,ne=te.maxMetadataBytes===!1?Number.POSITIVE_INFINITY:te.maxMetadataBytes,Oe=2*1024,ie="__ragChunk",ce="__ragSource",A="__ragText",Q="__ragHash",V="__ragChunks",B="__ragImportance",de="__ragModel",Ke=new Set([ie,V,Q,B,de,ce,A]),Ue=(e,i,m,w)=>{if(w===!1)return;const E=new TextEncoder().encode(JSON.stringify(e)).length;if(E<=w)return;const R=(typeof e[A]=="string"?new TextEncoder().encode(e[A]).length:0)*2>E?"lower `chunkSize` (it counts characters, not bytes — multibyte text costs up to 3 bytes each), or supply `textStore` to move chunk text out of metadata entirely":"attach less per-source `metadata`";throw new g("BAD_REQUEST",`@lunora/ai/rag: chunk ${String(i)} of "${m}" carries ${String(E)} bytes of metadata, over the store's ${String(w)}-byte per-vector ceiling — ${R}`)},Qe=(e,i,m)=>{if(m===!1)return;const w=new TextEncoder().encode(e).length;if(!(w<=m))throw new g("BAD_REQUEST",`@lunora/ai/rag: chunk id "${e}" for source "${i}" is ${String(w)} bytes, over the store's ${String(m)}-byte per-vector id ceiling — shorten the source id (hash long keys before indexing them) or shorten the \`namespace\`, which is prefixed onto every chunk id`)},Ve=/^[\w.-]{1,40}$/,ue=e=>e===void 0?"":`${encodeURIComponent(e)}#`,T=(e,i,m)=>`${ue(e)}${i}#${String(m)}`,re=(e,i)=>{const m=ue(i),w=m!==""&&e.startsWith(m)?e.slice(m.length):e,E=w.lastIndexOf("#"),S=E===-1?Number.NaN:Number(w.slice(E+1));return E===-1||!Number.isInteger(S)||S<0?{chunkIndex:0,sourceId:w}:{chunkIndex:S,sourceId:w.slice(0,E)}},Fe=async e=>Me(new TextEncoder().encode(e)),ze=(e,i)=>e.filter(m=>m.score>=i),oe=e=>{if(!e)return;const i=Object.entries(e).filter(([m])=>!Ke.has(m));return i.length>0?Object.fromEntries(i):void 0},se=new Set,Le=e=>{se.has(e)||(se.add(e),console.warn(`[@lunora/ai/rag] index "${e}" is used without a namespace — in a multi-tenant/sharded
+import{LunoraError as g,isLunoraError as ye}from"@lunora/errors";import{tool as xe,jsonSchema as Ie,embedMany as Se,embed as _e}from"ai";import{estimateModelCost as ke}from"./DEFAULT_MODEL_PRICES-Q8uxdiuV.mjs";import Te from"./fixedWindowChunks-C461ahRE.mjs";import{c as Ae,I as Ne}from"./concurrent-C6nqBv41.mjs";import{contentHash as Me}from"./contentHash-BIn6ECP8.mjs";import ee from"./hybridRank-DejmVw2I.mjs";import{VECTORIZE_CAPABILITIES as te,vectorizeStore as De}from"./VECTORIZE_CAPABILITIES-CUQDoxis.mjs";const Re=1e3,Ce=200,$e=5,Be=4,ne=te.maxMetadataBytes===!1?Number.POSITIVE_INFINITY:te.maxMetadataBytes,Oe=2*1024,ie="__ragChunk",ce="__ragSource",A="__ragText",Q="__ragHash",V="__ragChunks",B="__ragImportance",de="__ragModel",Ke=new Set([ie,V,Q,B,de,ce,A]),Ue=(e,i,m,w)=>{if(w===!1)return;const E=new TextEncoder().encode(JSON.stringify(e)).length;if(E<=w)return;const R=(typeof e[A]=="string"?new TextEncoder().encode(e[A]).length:0)*2>E?"lower `chunkSize` (it counts characters, not bytes — multibyte text costs up to 3 bytes each), or supply `textStore` to move chunk text out of metadata entirely":"attach less per-source `metadata`";throw new g("BAD_REQUEST",`@lunora/ai/rag: chunk ${String(i)} of "${m}" carries ${String(E)} bytes of metadata, over the store's ${String(w)}-byte per-vector ceiling — ${R}`)},Qe=(e,i,m)=>{if(m===!1)return;const w=new TextEncoder().encode(e).length;if(!(w<=m))throw new g("BAD_REQUEST",`@lunora/ai/rag: chunk id "${e}" for source "${i}" is ${String(w)} bytes, over the store's ${String(m)}-byte per-vector id ceiling — shorten the source id (hash long keys before indexing them) or shorten the \`namespace\`, which is prefixed onto every chunk id`)},Ve=/^[\w.-]{1,40}$/,ue=e=>e===void 0?"":`${encodeURIComponent(e)}#`,T=(e,i,m)=>`${ue(e)}${i}#${String(m)}`,re=(e,i)=>{const m=ue(i),w=m!==""&&e.startsWith(m)?e.slice(m.length):e,E=w.lastIndexOf("#"),S=E===-1?Number.NaN:Number(w.slice(E+1));return E===-1||!Number.isInteger(S)||S<0?{chunkIndex:0,sourceId:w}:{chunkIndex:S,sourceId:w.slice(0,E)}},Fe=async e=>Me(new TextEncoder().encode(e)),ze=(e,i)=>e.filter(m=>m.score>=i),oe=e=>{if(!e)return;const i=Object.entries(e).filter(([m])=>!Ke.has(m));return i.length>0?Object.fromEntries(i):void 0},se=new Set,Le=e=>{se.has(e)||(se.add(e),console.warn(`[@lunora/ai/rag] index "${e}" is used without a namespace — in a multi-tenant/sharded
 app this shares one tenant's chunks (text included) with every other tenant, since
 Vectorize indexes are account-global. Pass \`namespace\` (the shard/tenant key) on both
 index() and retrieve(). Single-tenant apps suppress this via { allowSharedNamespace: true }.`))},Pe=e=>e.map(i=>`[source:${i.sourceId}#${String(i.chunkIndex)}]

package/dist/packem_shared/sql-CNy5-Oop.mjs

@@ -0,0 +1 @@
+const l=t=>/^[A-Z_]\w*$/i.test(t),h=t=>Array.from({length:t}).fill("?").join(", "),f=64,d=(t,e=64)=>{const n=Math.max(1,Math.floor(e)),s=[];for(let r=0;r<t.length;r+=n)s.push(t.slice(r,r+n));return s},g=(t,e)=>{if(!l(t))throw new TypeError(`@lunora/ai/rag: ${e} must be a bare SQL identifier (letters, digits, underscore; not starting with a digit) — got "${t}"`);return t},u=(t,e)=>{if(t.length!==e.length||t.length===0)return 0;let n=0,s=0,r=0;for(const[c,i]of t.entries()){const o=e[c];n+=i*o,s+=i*i,r+=o*o}const a=Math.sqrt(s)*Math.sqrt(r);return a===0?0:n/a},p=t=>{if(t!=null){if(typeof t=="string")try{return JSON.parse(t)}catch{return}return t}};export{f as I,g as a,u as c,d as i,h as p,p as r};

package/dist/packem_shared/{sqlLexicalStore-DcWChCVZ.mjs → sqlLexicalStore-qkTJHHzM.mjs}

@@ -1 +1 @@
-import{t as x,c as U,b,a as F}from"./bm25-BT0cd0lY.mjs";import M from"./matchesMetadataFilter-BbIOyA5g.mjs";import{a as q,i as p,p as L,r as C,I as D}from"./sql-D6W_98IK.mjs";const X="lunora_rag_lexical",_=null,I=m=>m??"",S=m=>{const s=typeof m=="number"?m:Number(m);return Number.isFinite(s)?s:0},K=m=>{if(typeof m.exec!="function")throw new TypeError("@lunora/ai/rag: sqlLexicalStore requires an `exec` function");const s=q(m.table??X,"sqlLexicalStore `table`"),f=`${s}_terms`,{exec:r}=m;let w;const h=async()=>{w??=(async()=>{await r(`CREATE TABLE IF NOT EXISTS ${s} (id TEXT NOT NULL, namespace TEXT NOT NULL DEFAULT '', text TEXT NOT NULL, length INTEGER NOT NULL, metadata TEXT, PRIMARY KEY (namespace, id))`,[]),await r(`CREATE TABLE IF NOT EXISTS ${f} (term TEXT NOT NULL, id TEXT NOT NULL, namespace TEXT NOT NULL DEFAULT '', frequency INTEGER NOT NULL, PRIMARY KEY (namespace, term, id))`,[]),await r(`CREATE INDEX IF NOT EXISTS ${f}_lookup ON ${f} (namespace, term)`,[]),await r(`CREATE INDEX IF NOT EXISTS ${s}_namespace ON ${s} (namespace)`,[])})().catch(c=>{throw w=void 0,c}),await w},A=async(c,n)=>{for(const o of p(c)){const i=L(o.length);await r(`DELETE FROM ${f} WHERE namespace = ? AND id IN (${i})`,[n,...o]),await r(`DELETE FROM ${s} WHERE namespace = ? AND id IN (${i})`,[n,...o])}},R=async(c,n,o)=>{const i=`(${L(n)})`;for(const E of p(o,D/n))await r(`${c} VALUES ${Array.from({length:E.length}).fill(i).join(", ")}`,E.flatMap(e=>[...e]))};return{index:async(c,n)=>{await h();const o=I(n.namespace);await A(c.map(e=>e.id),o);const i=[],E=[];for(const e of c){const N=F(e.text);if(N.length===0)continue;const d=new Map;for(const T of N)d.set(T,(d.get(T)??0)+1);i.push([e.id,o,e.text,N.length,e.metadata===void 0?_:JSON.stringify(e.metadata)]);for(const[T,l]of d)E.push([T,e.id,o,l])}await R(`INSERT INTO ${f} (term, id, namespace, frequency)`,4,E),await R(`INSERT INTO ${s} (id, namespace, text, length, metadata)`,5,i)},remove:async(c,n)=>{await h(),await A(c,I(n.namespace))},search:async(c,n)=>{await h();const o=I(n.namespace),i=x(c);if(i.length===0)return[];const[E]=await r(`SELECT COUNT(*) AS document_count, COALESCE(SUM(length), 0) AS total_length FROM ${s} WHERE namespace = ?`,[o]),e=S(E?.document_count);if(e===0)return[];const N=S(E?.total_length)/e,d=[];for(const t of p(i)){const a=await r(`SELECT t.term AS term, t.id AS id, t.frequency AS frequency, d.length AS length, d.metadata AS metadata FROM ${f} t JOIN ${s} d ON d.id = t.id AND d.namespace = t.namespace WHERE t.namespace = ? AND t.term IN (${L(t.length)})`,[o,...t]);d.push(...a)}const T=new Map;for(const t of d){const a=String(t.term);T.set(a,(T.get(a)??0)+1)}const l=new Map;for(const t of d){const a=C(t.metadata);if(!M(a,n.filter))continue;const u=String(t.id),y=U(e,T.get(String(t.term))??1),$=b(y,S(t.frequency),S(t.length),N);l.set(u,(l.get(u)??0)+$)}const g=[...l.entries()].toSorted(([,t],[,a])=>a-t).slice(0,n.topK);if(g.length===0)return[];const O=new Map;for(const t of p(g.map(([a])=>a))){const a=await r(`SELECT id, text FROM ${s} WHERE namespace = ? AND id IN (${L(t.length)})`,[o,...t]);for(const u of a)O.set(String(u.id),String(u.text))}return g.map(([t,a])=>({id:t,score:a,text:O.get(t)??""}))}}};export{K as sqlLexicalStore};
+import{t as x,c as U,b,a as F}from"./bm25-BT0cd0lY.mjs";import M from"./matchesMetadataFilter-BbIOyA5g.mjs";import{a as q,i as p,p as L,r as C,I as D}from"./sql-CNy5-Oop.mjs";const X="lunora_rag_lexical",_=null,I=m=>m??"",S=m=>{const s=typeof m=="number"?m:Number(m);return Number.isFinite(s)?s:0},K=m=>{if(typeof m.exec!="function")throw new TypeError("@lunora/ai/rag: sqlLexicalStore requires an `exec` function");const s=q(m.table??X,"sqlLexicalStore `table`"),f=`${s}_terms`,{exec:r}=m;let w;const h=async()=>{w??=(async()=>{await r(`CREATE TABLE IF NOT EXISTS ${s} (id TEXT NOT NULL, namespace TEXT NOT NULL DEFAULT '', text TEXT NOT NULL, length INTEGER NOT NULL, metadata TEXT, PRIMARY KEY (namespace, id))`,[]),await r(`CREATE TABLE IF NOT EXISTS ${f} (term TEXT NOT NULL, id TEXT NOT NULL, namespace TEXT NOT NULL DEFAULT '', frequency INTEGER NOT NULL, PRIMARY KEY (namespace, term, id))`,[]),await r(`CREATE INDEX IF NOT EXISTS ${f}_lookup ON ${f} (namespace, term)`,[]),await r(`CREATE INDEX IF NOT EXISTS ${s}_namespace ON ${s} (namespace)`,[])})().catch(c=>{throw w=void 0,c}),await w},A=async(c,n)=>{for(const o of p(c)){const i=L(o.length);await r(`DELETE FROM ${f} WHERE namespace = ? AND id IN (${i})`,[n,...o]),await r(`DELETE FROM ${s} WHERE namespace = ? AND id IN (${i})`,[n,...o])}},R=async(c,n,o)=>{const i=`(${L(n)})`;for(const E of p(o,D/n))await r(`${c} VALUES ${Array.from({length:E.length}).fill(i).join(", ")}`,E.flatMap(e=>[...e]))};return{index:async(c,n)=>{await h();const o=I(n.namespace);await A(c.map(e=>e.id),o);const i=[],E=[];for(const e of c){const N=F(e.text);if(N.length===0)continue;const d=new Map;for(const T of N)d.set(T,(d.get(T)??0)+1);i.push([e.id,o,e.text,N.length,e.metadata===void 0?_:JSON.stringify(e.metadata)]);for(const[T,l]of d)E.push([T,e.id,o,l])}await R(`INSERT INTO ${f} (term, id, namespace, frequency)`,4,E),await R(`INSERT INTO ${s} (id, namespace, text, length, metadata)`,5,i)},remove:async(c,n)=>{await h(),await A(c,I(n.namespace))},search:async(c,n)=>{await h();const o=I(n.namespace),i=x(c);if(i.length===0)return[];const[E]=await r(`SELECT COUNT(*) AS document_count, COALESCE(SUM(length), 0) AS total_length FROM ${s} WHERE namespace = ?`,[o]),e=S(E?.document_count);if(e===0)return[];const N=S(E?.total_length)/e,d=[];for(const t of p(i)){const a=await r(`SELECT t.term AS term, t.id AS id, t.frequency AS frequency, d.length AS length, d.metadata AS metadata FROM ${f} t JOIN ${s} d ON d.id = t.id AND d.namespace = t.namespace WHERE t.namespace = ? AND t.term IN (${L(t.length)})`,[o,...t]);d.push(...a)}const T=new Map;for(const t of d){const a=String(t.term);T.set(a,(T.get(a)??0)+1)}const l=new Map;for(const t of d){const a=C(t.metadata);if(!M(a,n.filter))continue;const u=String(t.id),y=U(e,T.get(String(t.term))??1),$=b(y,S(t.frequency),S(t.length),N);l.set(u,(l.get(u)??0)+$)}const g=[...l.entries()].toSorted(([,t],[,a])=>a-t).slice(0,n.topK);if(g.length===0)return[];const O=new Map;for(const t of p(g.map(([a])=>a))){const a=await r(`SELECT id, text FROM ${s} WHERE namespace = ? AND id IN (${L(t.length)})`,[o,...t]);for(const u of a)O.set(String(u.id),String(u.text))}return g.map(([t,a])=>({id:t,score:a,text:O.get(t)??""}))}}};export{K as sqlLexicalStore};

package/dist/packem_shared/{sqliteVectorStore-B4HiD_-t.mjs → sqliteVectorStore-DEqrp-Uc.mjs}

@@ -1 +1 @@
-import g from"./matchesMetadataFilter-BbIOyA5g.mjs";import{a as v,p as f,r as h,c as L,i as S}from"./sql-D6W_98IK.mjs";const N="lunora_rag_vectors",p=5e4,w=100,I=null,d=s=>s??"",R=s=>{if(typeof s.exec!="function")throw new TypeError("@lunora/ai/rag: sqliteVectorStore requires an `exec` function");const n=v(s.table??N,"sqliteVectorStore `table`"),E=s.maxScan??p,{exec:c}=s,y={maxDimensions:s.maxDimensions??!1,maxIdBytes:!1,maxMetadataBytes:!1,maxTopK:w,maxTopKWithMetadata:w};let T;const m=async()=>{T??=(async()=>{await c(`CREATE TABLE IF NOT EXISTS ${n} (id TEXT NOT NULL, namespace TEXT NOT NULL DEFAULT '', vector TEXT NOT NULL, metadata TEXT, PRIMARY KEY (namespace, id))`,[]),await c(`CREATE INDEX IF NOT EXISTS ${n}_namespace ON ${n} (namespace)`,[])})().catch(e=>{throw T=void 0,e}),await T};return{capabilities:y,deleteByIds:async(e,t)=>{if(await m(),e.length!==0)for(const a of S(e))await c(`DELETE FROM ${n} WHERE namespace = ? AND id IN (${f(a.length)})`,[d(t),...a])},getByIds:async(e,t)=>{if(await m(),e.length===0)return[];const a=[];for(const i of S(e)){const l=await c(`SELECT id, metadata FROM ${n} WHERE namespace = ? AND id IN (${f(i.length)})`,[d(t),...i]);for(const r of l){const o=h(r.metadata);a.push({id:String(r.id),...o===void 0?{}:{metadata:o}})}}return a},query:async e=>{await m();let t;if(e.embed&&e.input!==void 0)t=await e.embed(e.input);else throw new TypeError("@lunora/ai/rag: sqliteVectorStore query requires both `input` and `embed`");const a=await c(`SELECT id, vector, metadata FROM ${n} WHERE namespace = ? LIMIT ?`,[d(e.namespace),E+1]);if(a.length>E)throw new RangeError(`@lunora/ai/rag: sqliteVectorStore scanned ${String(a.length)} vectors in namespace "${d(e.namespace)}", over the ${String(E)} limit — search here is brute force and linear, so this namespace has outgrown it. Shard it further, or move this index to Vectorize or a pgvector backend`);const i=[];for(const r of a){const o=h(r.metadata);if(!g(o,e.filter))continue;const u=h(r.vector);u!==void 0&&i.push({id:String(r.id),score:L(t,u),...e.returnMetadata==="none"||o===void 0?{}:{metadata:o}})}const l=i.toSorted((r,o)=>o.score-r.score).slice(0,e.topK??10);return{count:l.length,matches:l}},upsert:async e=>{if(await m(),!e.embed)throw new TypeError("@lunora/ai/rag: sqliteVectorStore requires an `embed` function on upsert");const t=await e.embed(e.input);await c(`INSERT INTO ${n} (id, namespace, vector, metadata) VALUES (${f(4)}) ON CONFLICT(namespace, id) DO UPDATE SET vector = excluded.vector, metadata = excluded.metadata`,[e.id,d(e.namespace),JSON.stringify([...t]),e.metadata===void 0?I:JSON.stringify(e.metadata)])}}};export{R as sqliteVectorStore};
+import g from"./matchesMetadataFilter-BbIOyA5g.mjs";import{a as v,p as f,r as h,c as L,i as S}from"./sql-CNy5-Oop.mjs";const N="lunora_rag_vectors",p=5e4,w=100,I=null,d=s=>s??"",R=s=>{if(typeof s.exec!="function")throw new TypeError("@lunora/ai/rag: sqliteVectorStore requires an `exec` function");const n=v(s.table??N,"sqliteVectorStore `table`"),E=s.maxScan??p,{exec:c}=s,y={maxDimensions:s.maxDimensions??!1,maxIdBytes:!1,maxMetadataBytes:!1,maxTopK:w,maxTopKWithMetadata:w};let T;const m=async()=>{T??=(async()=>{await c(`CREATE TABLE IF NOT EXISTS ${n} (id TEXT NOT NULL, namespace TEXT NOT NULL DEFAULT '', vector TEXT NOT NULL, metadata TEXT, PRIMARY KEY (namespace, id))`,[]),await c(`CREATE INDEX IF NOT EXISTS ${n}_namespace ON ${n} (namespace)`,[])})().catch(e=>{throw T=void 0,e}),await T};return{capabilities:y,deleteByIds:async(e,t)=>{if(await m(),e.length!==0)for(const a of S(e))await c(`DELETE FROM ${n} WHERE namespace = ? AND id IN (${f(a.length)})`,[d(t),...a])},getByIds:async(e,t)=>{if(await m(),e.length===0)return[];const a=[];for(const i of S(e)){const l=await c(`SELECT id, metadata FROM ${n} WHERE namespace = ? AND id IN (${f(i.length)})`,[d(t),...i]);for(const r of l){const o=h(r.metadata);a.push({id:String(r.id),...o===void 0?{}:{metadata:o}})}}return a},query:async e=>{await m();let t;if(e.embed&&e.input!==void 0)t=await e.embed(e.input);else throw new TypeError("@lunora/ai/rag: sqliteVectorStore query requires both `input` and `embed`");const a=await c(`SELECT id, vector, metadata FROM ${n} WHERE namespace = ? LIMIT ?`,[d(e.namespace),E+1]);if(a.length>E)throw new RangeError(`@lunora/ai/rag: sqliteVectorStore scanned ${String(a.length)} vectors in namespace "${d(e.namespace)}", over the ${String(E)} limit — search here is brute force and linear, so this namespace has outgrown it. Shard it further, or move this index to Vectorize or a pgvector backend`);const i=[];for(const r of a){const o=h(r.metadata);if(!g(o,e.filter))continue;const u=h(r.vector);u!==void 0&&i.push({id:String(r.id),score:L(t,u),...e.returnMetadata==="none"||o===void 0?{}:{metadata:o}})}const l=i.toSorted((r,o)=>o.score-r.score).slice(0,e.topK??10);return{count:l.length,matches:l}},upsert:async e=>{if(await m(),!e.embed)throw new TypeError("@lunora/ai/rag: sqliteVectorStore requires an `embed` function on upsert");const t=await e.embed(e.input);await c(`INSERT INTO ${n} (id, namespace, vector, metadata) VALUES (${f(4)}) ON CONFLICT(namespace, id) DO UPDATE SET vector = excluded.vector, metadata = excluded.metadata`,[e.id,d(e.namespace),JSON.stringify([...t]),e.metadata===void 0?I:JSON.stringify(e.metadata)])}}};export{R as sqliteVectorStore};

package/dist/rag/index.d.mts

@@ -1026,18 +1026,6 @@ interface RagSourceSync {
  * @experimental
  */
 declare const defineRagSource: (rag: Rag, options?: RagSourceOptions) => RagSourceSync;
-/**
- * The injected SQL seam the durable RAG stores share.
- *
- * `@lunora/ai` depends on no database. A store is handed an executor instead,
- * so the same adapter runs on a Durable Object's SQLite, D1, `node:sqlite`, or
- * Postgres over Hyperdrive — and this package stays free of every one of them.
- *
- * The statements the stores emit are SQLite-shaped (`ON CONFLICT … DO UPDATE`,
- * `CREATE INDEX IF NOT EXISTS`, `?` placeholders), so the executor must front a
- * SQLite engine — a Durable Object, D1, or `node:sqlite`.
- * @experimental
- */
 /**
  * Run one statement and return its rows.
  *

package/dist/rag/index.d.ts

@@ -1026,18 +1026,6 @@ interface RagSourceSync {
  * @experimental
  */
 declare const defineRagSource: (rag: Rag, options?: RagSourceOptions) => RagSourceSync;
-/**
- * The injected SQL seam the durable RAG stores share.
- *
- * `@lunora/ai` depends on no database. A store is handed an executor instead,
- * so the same adapter runs on a Durable Object's SQLite, D1, `node:sqlite`, or
- * Postgres over Hyperdrive — and this package stays free of every one of them.
- *
- * The statements the stores emit are SQLite-shaped (`ON CONFLICT … DO UPDATE`,
- * `CREATE INDEX IF NOT EXISTS`, `?` placeholders), so the executor must front a
- * SQLite engine — a Durable Object, D1, or `node:sqlite`.
- * @experimental
- */
 /**
  * Run one statement and return its rows.
  *

package/dist/rag/index.mjs

@@ -1 +1 @@
-import{default as o}from"../packem_shared/fixedWindowChunks-C461ahRE.mjs";import{markdownChunker as a,sentenceChunker as f,tokenChunker as n}from"../packem_shared/markdownChunker-CDME5gEs.mjs";import{default as x}from"../packem_shared/defineRag-Dzh-MbCu.mjs";import{contentHash as p,guessMimeTypeFromExtension as i}from"../packem_shared/contentHash-BIn6ECP8.mjs";import{default as c}from"../packem_shared/hybridRank-DejmVw2I.mjs";import{default as l}from"../packem_shared/bm25LexicalStore-CALA7Lhm.mjs";import{default as h}from"../packem_shared/matchesMetadataFilter-BbIOyA5g.mjs";import{batchReranker as g,scoreReranker as C}from"../packem_shared/batchReranker-Bc38FBLH.mjs";import{defineRagSource as E}from"../packem_shared/defineRagSource-Q3f3niU8.mjs";import{sqlLexicalStore as T}from"../packem_shared/sqlLexicalStore-DcWChCVZ.mjs";import{sqliteVectorStore as y}from"../packem_shared/sqliteVectorStore-B4HiD_-t.mjs";import{ragSyncTriggers as q}from"../packem_shared/ragSyncTriggers-DPqzBNFw.mjs";import{VECTORIZE_CAPABILITIES as A,vectorizeStore as F}from"../packem_shared/VECTORIZE_CAPABILITIES-CUQDoxis.mjs";export{A as VECTORIZE_CAPABILITIES,g as batchReranker,l as bm25LexicalStore,p as contentHash,x as defineRag,E as defineRagSource,o as fixedWindowChunks,i as guessMimeTypeFromExtension,c as hybridRank,a as markdownChunker,h as matchesMetadataFilter,q as ragSyncTriggers,C as scoreReranker,f as sentenceChunker,T as sqlLexicalStore,y as sqliteVectorStore,n as tokenChunker,F as vectorizeStore};
+import{default as o}from"../packem_shared/fixedWindowChunks-C461ahRE.mjs";import{markdownChunker as a,sentenceChunker as f,tokenChunker as n}from"../packem_shared/markdownChunker-CDME5gEs.mjs";import{default as x}from"../packem_shared/defineRag-CMM17YeT.mjs";import{contentHash as p,guessMimeTypeFromExtension as i}from"../packem_shared/contentHash-BIn6ECP8.mjs";import{default as c}from"../packem_shared/hybridRank-DejmVw2I.mjs";import{default as l}from"../packem_shared/bm25LexicalStore-CALA7Lhm.mjs";import{default as h}from"../packem_shared/matchesMetadataFilter-BbIOyA5g.mjs";import{batchReranker as g,scoreReranker as C}from"../packem_shared/batchReranker-Bc38FBLH.mjs";import{defineRagSource as E}from"../packem_shared/defineRagSource-Q3f3niU8.mjs";import{sqlLexicalStore as T}from"../packem_shared/sqlLexicalStore-qkTJHHzM.mjs";import{sqliteVectorStore as y}from"../packem_shared/sqliteVectorStore-DEqrp-Uc.mjs";import{ragSyncTriggers as q}from"../packem_shared/ragSyncTriggers-DPqzBNFw.mjs";import{VECTORIZE_CAPABILITIES as A,vectorizeStore as F}from"../packem_shared/VECTORIZE_CAPABILITIES-CUQDoxis.mjs";export{A as VECTORIZE_CAPABILITIES,g as batchReranker,l as bm25LexicalStore,p as contentHash,x as defineRag,E as defineRagSource,o as fixedWindowChunks,i as guessMimeTypeFromExtension,c as hybridRank,a as markdownChunker,h as matchesMetadataFilter,q as ragSyncTriggers,C as scoreReranker,f as sentenceChunker,T as sqlLexicalStore,y as sqliteVectorStore,n as tokenChunker,F as vectorizeStore};

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@lunora/ai",
-  "version": "1.0.0-alpha.64",
+  "version": "1.0.0-alpha.66",
   "description": "Workers AI helper for Lunora: provider-agnostic AI SDK access from functions, Workers AI by default",
   "keywords": [
     "ai",

package/dist/packem_shared/DEFAULT_MODEL_PRICES-DZ3dYVUn.mjs

@@ -1 +0,0 @@
-const r=/^(.*)-\d{4}-\d{2}-\d{2}$/u,a={"@cf/baai/bge-base-en-v1.5":{input:.067},"@cf/baai/bge-large-en-v1.5":{input:.204},"@cf/baai/bge-m3":{input:.012},"@cf/baai/bge-small-en-v1.5":{input:.02},"text-embedding-3-large":{input:.13},"text-embedding-3-small":{input:.02}},c=n=>{const t=n.trim(),e=t.lastIndexOf("/");return(e!==-1&&!t.startsWith("@")?[t,t.slice(e+1)]:[t]).flatMap(i=>{const o=r.exec(i)?.[1];return o===void 0?[i]:[i,o]})},b=(n,t=a)=>{for(const e of c(n))if(Object.hasOwn(t,e))return t[e]},d=(n,t,e)=>{if(n===void 0||n.length===0)return;const s=b(n,e);if(s===void 0)return;const i=Number.isFinite(t.inputTokens)?Math.max(0,t.inputTokens):0,o=Number.isFinite(t.outputTokens)?Math.max(0,t.outputTokens):0;if(i<=0&&o<=0)return;const u=(i*s.input+o*(s.output??0))/1e6;return Number.isFinite(u)?u:void 0};export{a as DEFAULT_MODEL_PRICES,d as estimateModelCost,b as lookupModelPrice};

package/dist/packem_shared/sql-D6W_98IK.mjs

@@ -1 +0,0 @@
-const h=t=>Array.from({length:t}).fill("?").join(", "),f=64,d=(t,e=64)=>{const n=Math.max(1,Math.floor(e)),s=[];for(let r=0;r<t.length;r+=n)s.push(t.slice(r,r+n));return s},l=/^[A-Z_]\w*$/i,g=(t,e)=>{if(!l.test(t))throw new TypeError(`@lunora/ai/rag: ${e} must be a bare SQL identifier (letters, digits, underscore; not starting with a digit) — got "${t}"`);return t},u=(t,e)=>{if(t.length!==e.length||t.length===0)return 0;let n=0,s=0,r=0;for(const[c,o]of t.entries()){const i=e[c];n+=o*i,s+=o*o,r+=i*i}const a=Math.sqrt(s)*Math.sqrt(r);return a===0?0:n/a},I=t=>{if(t!=null){if(typeof t=="string")try{return JSON.parse(t)}catch{return}return t}};export{f as I,g as a,u as c,d as i,h as p,I as r};