@ariaflowagents/rag 0.4.3
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/README.md +5 -0
- package/dist/chunkers.d.ts +4 -0
- package/dist/chunkers.d.ts.map +1 -0
- package/dist/chunkers.js +77 -0
- package/dist/chunkers.js.map +1 -0
- package/dist/index.d.ts +5 -0
- package/dist/index.d.ts.map +1 -0
- package/dist/index.js +4 -0
- package/dist/index.js.map +1 -0
- package/dist/llmRetriever.d.ts +3 -0
- package/dist/llmRetriever.d.ts.map +1 -0
- package/dist/llmRetriever.js +88 -0
- package/dist/llmRetriever.js.map +1 -0
- package/dist/source.d.ts +13 -0
- package/dist/source.d.ts.map +1 -0
- package/dist/source.js +17 -0
- package/dist/source.js.map +1 -0
- package/dist/types.d.ts +40 -0
- package/dist/types.d.ts.map +1 -0
- package/dist/types.js +2 -0
- package/dist/types.js.map +1 -0
- package/package.json +36 -0
package/README.md
ADDED
|
@@ -0,0 +1,4 @@
|
|
|
1
|
+
import type { Chunker, ChunkOptions } from './types.js';
|
|
2
|
+
export declare function createMarkdownChunker(defaultOptions?: ChunkOptions): Chunker;
|
|
3
|
+
export declare function createRecursiveChunker(defaultOptions?: ChunkOptions): Chunker;
|
|
4
|
+
//# sourceMappingURL=chunkers.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"chunkers.d.ts","sourceRoot":"","sources":["../src/chunkers.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,OAAO,EAAE,YAAY,EAAkB,MAAM,YAAY,CAAC;AAExE,wBAAgB,qBAAqB,CAAC,cAAc,GAAE,YAAiB,GAAG,OAAO,CA0BhF;AAED,wBAAgB,sBAAsB,CAAC,cAAc,GAAE,YAAiB,GAAG,OAAO,CAQjF"}
|
package/dist/chunkers.js
ADDED
|
@@ -0,0 +1,77 @@
|
|
|
1
|
+
export function createMarkdownChunker(defaultOptions = {}) {
|
|
2
|
+
return {
|
|
3
|
+
chunk(text, options) {
|
|
4
|
+
const merged = { ...defaultOptions, ...options };
|
|
5
|
+
const sections = parseMarkdownSections(text);
|
|
6
|
+
const chunks = [];
|
|
7
|
+
const seen = new Map();
|
|
8
|
+
for (const section of sections) {
|
|
9
|
+
const baseId = section.id;
|
|
10
|
+
const count = (seen.get(baseId) ?? 0) + 1;
|
|
11
|
+
seen.set(baseId, count);
|
|
12
|
+
const id = count > 1 ? `${baseId} (${count})` : baseId;
|
|
13
|
+
const parts = splitByMaxChars(section.text, merged.maxChars, merged.overlapChars);
|
|
14
|
+
if (parts.length === 1) {
|
|
15
|
+
chunks.push({ id, text: parts[0] });
|
|
16
|
+
continue;
|
|
17
|
+
}
|
|
18
|
+
for (let i = 0; i < parts.length; i++) {
|
|
19
|
+
chunks.push({ id: `${id}#${i + 1}`, text: parts[i] });
|
|
20
|
+
}
|
|
21
|
+
}
|
|
22
|
+
return chunks;
|
|
23
|
+
},
|
|
24
|
+
};
|
|
25
|
+
}
|
|
26
|
+
export function createRecursiveChunker(defaultOptions = {}) {
|
|
27
|
+
return {
|
|
28
|
+
chunk(text, options) {
|
|
29
|
+
const merged = { ...defaultOptions, ...options };
|
|
30
|
+
const parts = splitByMaxChars(text, merged.maxChars, merged.overlapChars);
|
|
31
|
+
return parts.map((part, index) => ({ id: `chunk-${index + 1}`, text: part }));
|
|
32
|
+
},
|
|
33
|
+
};
|
|
34
|
+
}
|
|
35
|
+
function parseMarkdownSections(content) {
|
|
36
|
+
const lines = content.split(/\r?\n/);
|
|
37
|
+
const sections = [];
|
|
38
|
+
let currentId = null;
|
|
39
|
+
let buffer = [];
|
|
40
|
+
let firstHeadingUsed = false;
|
|
41
|
+
for (const line of lines) {
|
|
42
|
+
const h1 = line.match(/^#\s+(.+)$/);
|
|
43
|
+
const hN = line.match(/^#{2,6}\s+(.+)$/);
|
|
44
|
+
if (h1 && !firstHeadingUsed && buffer.length === 0 && !currentId) {
|
|
45
|
+
firstHeadingUsed = true;
|
|
46
|
+
continue;
|
|
47
|
+
}
|
|
48
|
+
if (hN) {
|
|
49
|
+
if (currentId || buffer.length > 0) {
|
|
50
|
+
sections.push({ id: currentId ?? 'section', text: buffer.join('\n').trim() });
|
|
51
|
+
}
|
|
52
|
+
currentId = hN[1].trim();
|
|
53
|
+
buffer = [];
|
|
54
|
+
continue;
|
|
55
|
+
}
|
|
56
|
+
buffer.push(line);
|
|
57
|
+
}
|
|
58
|
+
if (currentId || buffer.length > 0) {
|
|
59
|
+
sections.push({ id: currentId ?? 'section', text: buffer.join('\n').trim() });
|
|
60
|
+
}
|
|
61
|
+
return sections.filter(section => section.text.length > 0);
|
|
62
|
+
}
|
|
63
|
+
function splitByMaxChars(text, maxChars, overlapChars = 0) {
|
|
64
|
+
if (!maxChars || text.length <= maxChars) {
|
|
65
|
+
return [text];
|
|
66
|
+
}
|
|
67
|
+
const step = Math.max(1, maxChars - overlapChars);
|
|
68
|
+
const parts = [];
|
|
69
|
+
for (let start = 0; start < text.length; start += step) {
|
|
70
|
+
const end = Math.min(text.length, start + maxChars);
|
|
71
|
+
parts.push(text.slice(start, end));
|
|
72
|
+
if (end === text.length)
|
|
73
|
+
break;
|
|
74
|
+
}
|
|
75
|
+
return parts;
|
|
76
|
+
}
|
|
77
|
+
//# sourceMappingURL=chunkers.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"chunkers.js","sourceRoot":"","sources":["../src/chunkers.ts"],"names":[],"mappings":"AAEA,MAAM,UAAU,qBAAqB,CAAC,iBAA+B,EAAE;IACrE,OAAO;QACL,KAAK,CAAC,IAAY,EAAE,OAAsB;YACxC,MAAM,MAAM,GAAG,EAAE,GAAG,cAAc,EAAE,GAAG,OAAO,EAAE,CAAC;YACjD,MAAM,QAAQ,GAAG,qBAAqB,CAAC,IAAI,CAAC,CAAC;YAC7C,MAAM,MAAM,GAAqB,EAAE,CAAC;YACpC,MAAM,IAAI,GAAG,IAAI,GAAG,EAAkB,CAAC;YAEvC,KAAK,MAAM,OAAO,IAAI,QAAQ,EAAE,CAAC;gBAC/B,MAAM,MAAM,GAAG,OAAO,CAAC,EAAE,CAAC;gBAC1B,MAAM,KAAK,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,MAAM,CAAC,IAAI,CAAC,CAAC,GAAG,CAAC,CAAC;gBAC1C,IAAI,CAAC,GAAG,CAAC,MAAM,EAAE,KAAK,CAAC,CAAC;gBACxB,MAAM,EAAE,GAAG,KAAK,GAAG,CAAC,CAAC,CAAC,CAAC,GAAG,MAAM,KAAK,KAAK,GAAG,CAAC,CAAC,CAAC,MAAM,CAAC;gBACvD,MAAM,KAAK,GAAG,eAAe,CAAC,OAAO,CAAC,IAAI,EAAE,MAAM,CAAC,QAAQ,EAAE,MAAM,CAAC,YAAY,CAAC,CAAC;gBAClF,IAAI,KAAK,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;oBACvB,MAAM,CAAC,IAAI,CAAC,EAAE,EAAE,EAAE,IAAI,EAAE,KAAK,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC;oBACpC,SAAS;gBACX,CAAC;gBACD,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,KAAK,CAAC,MAAM,EAAE,CAAC,EAAE,EAAE,CAAC;oBACtC,MAAM,CAAC,IAAI,CAAC,EAAE,EAAE,EAAE,GAAG,EAAE,IAAI,CAAC,GAAG,CAAC,EAAE,EAAE,IAAI,EAAE,KAAK,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC;gBACxD,CAAC;YACH,CAAC;YAED,OAAO,MAAM,CAAC;QAChB,CAAC;KACF,CAAC;AACJ,CAAC;AAED,MAAM,UAAU,sBAAsB,CAAC,iBAA+B,EAAE;IACtE,OAAO;QACL,KAAK,CAAC,IAAY,EAAE,OAAsB;YACxC,MAAM,MAAM,GAAG,EAAE,GAAG,cAAc,EAAE,GAAG,OAAO,EAAE,CAAC;YACjD,MAAM,KAAK,GAAG,eAAe,CAAC,IAAI,EAAE,MAAM,CAAC,QAAQ,EAAE,MAAM,CAAC,YAAY,CAAC,CAAC;YAC1E,OAAO,KAAK,CAAC,GAAG,CAAC,CAAC,IAAI,EAAE,KAAK,EAAE,EAAE,CAAC,CAAC,EAAE,EAAE,EAAE,SAAS,KAAK,GAAG,CAAC,EAAE,EAAE,IAAI,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC;QAChF,CAAC;KACF,CAAC;AACJ,CAAC;AAED,SAAS,qBAAqB,CAAC,OAAe;IAC5C,MAAM,KAAK,GAAG,OAAO,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC;IACrC,MAAM,QAAQ,GAAqB,EAAE,CAAC;IAEtC,IAAI,SAAS,GAAkB,IAAI,CAAC;IACpC,IAAI,MAAM,GAAa,EAAE,CAAC;IAC1B,IAAI,gBAAgB,GAAG,KAAK,CAAC;IAE7B,KAAK,MAAM,IAAI,IAAI,KAAK,EAAE,CAAC;QACzB,MAAM,EAAE,GAAG,IAAI,CAAC,KAAK,CAAC,YAAY,CAAC,CAAC;QACpC,MAAM,EAAE,GAAG,IAAI,CAAC,KAAK,CAAC,iBAAiB,CAAC,CAAC;QAEzC,IAAI,EAAE,IAAI,CAAC,gBAAgB,IAAI,MAAM,CAAC,MAAM,KAAK,CAAC,IAAI,CAAC,SAAS,EAAE,CAAC;YACjE,gBAAgB,GAAG,IAAI,CAAC;YACxB,SAAS;QACX,CAAC;QAED,IAAI,EAAE,EAAE,CAAC;YACP,IAAI,SAAS,IAAI,MAAM,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;gBACnC,QAAQ,CAAC,IAAI,CAAC,EAAE,EAAE,EAAE,SAAS,IAAI,SAAS,EAAE,IAAI,EAAE,MAAM,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC;YAChF,CAAC;YACD,SAAS,GAAG,EAAE,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC;YACzB,MAAM,GAAG,EAAE,CAAC;YACZ,SAAS;QACX,CAAC;QAED,MAAM,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;IACpB,CAAC;IAED,IAAI,SAAS,IAAI,MAAM,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;QACnC,QAAQ,CAAC,IAAI,CAAC,EAAE,EAAE,EAAE,SAAS,IAAI,SAAS,EAAE,IAAI,EAAE,MAAM,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC;IAChF,CAAC;IAED,OAAO,QAAQ,CAAC,MAAM,CAAC,OAAO,CAAC,EAAE,CAAC,OAAO,CAAC,IAAI,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC;AAC7D,CAAC;AAED,SAAS,eAAe,CAAC,IAAY,EAAE,QAAiB,EAAE,eAAuB,CAAC;IAChF,IAAI,CAAC,QAAQ,IAAI,IAAI,CAAC,MAAM,IAAI,QAAQ,EAAE,CAAC;QACzC,OAAO,CAAC,IAAI,CAAC,CAAC;IAChB,CAAC;IAED,MAAM,IAAI,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,QAAQ,GAAG,YAAY,CAAC,CAAC;IAClD,MAAM,KAAK,GAAa,EAAE,CAAC;IAE3B,KAAK,IAAI,KAAK,GAAG,CAAC,EAAE,KAAK,GAAG,IAAI,CAAC,MAAM,EAAE,KAAK,IAAI,IAAI,EAAE,CAAC;QACvD,MAAM,GAAG,GAAG,IAAI,CAAC,GAAG,CAAC,IAAI,CAAC,MAAM,EAAE,KAAK,GAAG,QAAQ,CAAC,CAAC;QACpD,KAAK,CAAC,IAAI,CAAC,IAAI,CAAC,KAAK,CAAC,KAAK,EAAE,GAAG,CAAC,CAAC,CAAC;QACnC,IAAI,GAAG,KAAK,IAAI,CAAC,MAAM;YAAE,MAAM;IACjC,CAAC;IAED,OAAO,KAAK,CAAC;AACf,CAAC"}
|
package/dist/index.d.ts
ADDED
|
@@ -0,0 +1,5 @@
|
|
|
1
|
+
export type { KnowledgeChunk, KnowledgeSource, Chunker, ChunkOptions, RetrievalHit, KnowledgeRetriever, LLMRetrieverOptions, } from './types.js';
|
|
2
|
+
export { createMarkdownChunker, createRecursiveChunker } from './chunkers.js';
|
|
3
|
+
export { createStaticKnowledgeSource } from './source.js';
|
|
4
|
+
export { createLLMRetriever } from './llmRetriever.js';
|
|
5
|
+
//# sourceMappingURL=index.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAAA,YAAY,EACV,cAAc,EACd,eAAe,EACf,OAAO,EACP,YAAY,EACZ,YAAY,EACZ,kBAAkB,EAClB,mBAAmB,GACpB,MAAM,YAAY,CAAC;AAEpB,OAAO,EAAE,qBAAqB,EAAE,sBAAsB,EAAE,MAAM,eAAe,CAAC;AAC9E,OAAO,EAAE,2BAA2B,EAAE,MAAM,aAAa,CAAC;AAC1D,OAAO,EAAE,kBAAkB,EAAE,MAAM,mBAAmB,CAAC"}
|
package/dist/index.js
ADDED
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"index.js","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAUA,OAAO,EAAE,qBAAqB,EAAE,sBAAsB,EAAE,MAAM,eAAe,CAAC;AAC9E,OAAO,EAAE,2BAA2B,EAAE,MAAM,aAAa,CAAC;AAC1D,OAAO,EAAE,kBAAkB,EAAE,MAAM,mBAAmB,CAAC"}
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"llmRetriever.d.ts","sourceRoot":"","sources":["../src/llmRetriever.ts"],"names":[],"mappings":"AAEA,OAAO,KAAK,EAAE,kBAAkB,EAAmB,mBAAmB,EAAgB,MAAM,YAAY,CAAC;AAYzG,wBAAgB,kBAAkB,CAAC,OAAO,EAAE,mBAAmB,GAAG,kBAAkB,CA4BnF"}
|
|
@@ -0,0 +1,88 @@
|
|
|
1
|
+
import { generateText, Output } from 'ai';
|
|
2
|
+
import { z } from 'zod';
|
|
3
|
+
const outputSchema = z.object({
|
|
4
|
+
ranked: z.array(z.object({
|
|
5
|
+
sourceId: z.string(),
|
|
6
|
+
chunkId: z.string(),
|
|
7
|
+
reason: z.string(),
|
|
8
|
+
})),
|
|
9
|
+
});
|
|
10
|
+
export function createLLMRetriever(options) {
|
|
11
|
+
const { model, topK = 4, includeReasons = true, candidateMaxChars = 1500 } = options;
|
|
12
|
+
return {
|
|
13
|
+
async retrieve(query, sources, opts) {
|
|
14
|
+
const k = Math.max(1, opts?.topK ?? topK);
|
|
15
|
+
const candidates = flattenCandidates(sources);
|
|
16
|
+
if (candidates.length === 0)
|
|
17
|
+
return [];
|
|
18
|
+
const prompt = buildPrompt({
|
|
19
|
+
query,
|
|
20
|
+
hint: opts?.hint,
|
|
21
|
+
candidates,
|
|
22
|
+
candidateMaxChars,
|
|
23
|
+
topK: k,
|
|
24
|
+
});
|
|
25
|
+
const { output } = await generateText({
|
|
26
|
+
model,
|
|
27
|
+
system: 'You are a retrieval ranker. Only select from provided chunk IDs.',
|
|
28
|
+
prompt,
|
|
29
|
+
output: Output.object({ schema: outputSchema }),
|
|
30
|
+
});
|
|
31
|
+
const ranked = output.ranked ?? [];
|
|
32
|
+
return coerceHits(ranked, candidates, k, includeReasons);
|
|
33
|
+
},
|
|
34
|
+
};
|
|
35
|
+
}
|
|
36
|
+
function flattenCandidates(sources) {
|
|
37
|
+
const candidates = [];
|
|
38
|
+
for (const source of sources) {
|
|
39
|
+
const chunks = source.getChunks();
|
|
40
|
+
for (const chunk of chunks) {
|
|
41
|
+
candidates.push({
|
|
42
|
+
sourceId: source.id,
|
|
43
|
+
chunkId: chunk.id,
|
|
44
|
+
text: chunk.text,
|
|
45
|
+
sourceName: source.name,
|
|
46
|
+
});
|
|
47
|
+
}
|
|
48
|
+
}
|
|
49
|
+
return candidates;
|
|
50
|
+
}
|
|
51
|
+
function buildPrompt(input) {
|
|
52
|
+
const header = [
|
|
53
|
+
`Query: ${input.query}`,
|
|
54
|
+
input.hint ? `Hint: ${input.hint}` : null,
|
|
55
|
+
`Return top ${input.topK} chunk IDs with short reasons.`,
|
|
56
|
+
'Only use provided sourceId + chunkId pairs.',
|
|
57
|
+
]
|
|
58
|
+
.filter(Boolean)
|
|
59
|
+
.join('\n');
|
|
60
|
+
const chunks = input.candidates
|
|
61
|
+
.map(candidate => {
|
|
62
|
+
const body = candidate.text.slice(0, input.candidateMaxChars);
|
|
63
|
+
return `sourceId: ${candidate.sourceId}\nchunkId: ${candidate.chunkId}\nsourceName: ${candidate.sourceName}\n${body}`;
|
|
64
|
+
})
|
|
65
|
+
.join('\n\n---\n\n');
|
|
66
|
+
return `${header}\n\nChunks:\n${chunks}`;
|
|
67
|
+
}
|
|
68
|
+
function coerceHits(ranked, candidates, topK, includeReasons) {
|
|
69
|
+
const candidateSet = new Set(candidates.map(c => `${c.sourceId}::${c.chunkId}`));
|
|
70
|
+
const hits = [];
|
|
71
|
+
for (const entry of ranked) {
|
|
72
|
+
const key = `${entry.sourceId}::${entry.chunkId}`;
|
|
73
|
+
if (!candidateSet.has(key))
|
|
74
|
+
continue;
|
|
75
|
+
if (hits.find(hit => hit.sourceId === entry.sourceId && hit.chunkId === entry.chunkId))
|
|
76
|
+
continue;
|
|
77
|
+
hits.push({
|
|
78
|
+
sourceId: entry.sourceId,
|
|
79
|
+
chunkId: entry.chunkId,
|
|
80
|
+
rank: hits.length + 1,
|
|
81
|
+
reason: includeReasons ? entry.reason ?? '' : undefined,
|
|
82
|
+
});
|
|
83
|
+
if (hits.length >= topK)
|
|
84
|
+
break;
|
|
85
|
+
}
|
|
86
|
+
return hits;
|
|
87
|
+
}
|
|
88
|
+
//# sourceMappingURL=llmRetriever.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"llmRetriever.js","sourceRoot":"","sources":["../src/llmRetriever.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,EAAE,MAAM,IAAI,CAAC;AAC1C,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAC;AAGxB,MAAM,YAAY,GAAG,CAAC,CAAC,MAAM,CAAC;IAC5B,MAAM,EAAE,CAAC,CAAC,KAAK,CACb,CAAC,CAAC,MAAM,CAAC;QACP,QAAQ,EAAE,CAAC,CAAC,MAAM,EAAE;QACpB,OAAO,EAAE,CAAC,CAAC,MAAM,EAAE;QACnB,MAAM,EAAE,CAAC,CAAC,MAAM,EAAE;KACnB,CAAC,CACH;CACF,CAAC,CAAC;AAEH,MAAM,UAAU,kBAAkB,CAAC,OAA4B;IAC7D,MAAM,EAAE,KAAK,EAAE,IAAI,GAAG,CAAC,EAAE,cAAc,GAAG,IAAI,EAAE,iBAAiB,GAAG,IAAI,EAAE,GAAG,OAAO,CAAC;IAErF,OAAO;QACL,KAAK,CAAC,QAAQ,CAAC,KAAa,EAAE,OAA0B,EAAE,IAAuC;YAC/F,MAAM,CAAC,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,EAAE,IAAI,IAAI,IAAI,CAAC,CAAC;YAC1C,MAAM,UAAU,GAAG,iBAAiB,CAAC,OAAO,CAAC,CAAC;YAC9C,IAAI,UAAU,CAAC,MAAM,KAAK,CAAC;gBAAE,OAAO,EAAE,CAAC;YAEvC,MAAM,MAAM,GAAG,WAAW,CAAC;gBACzB,KAAK;gBACL,IAAI,EAAE,IAAI,EAAE,IAAI;gBAChB,UAAU;gBACV,iBAAiB;gBACjB,IAAI,EAAE,CAAC;aACR,CAAC,CAAC;YAEH,MAAM,EAAE,MAAM,EAAE,GAAG,MAAM,YAAY,CAAC;gBACpC,KAAK;gBACL,MAAM,EAAE,kEAAkE;gBAC1E,MAAM;gBACN,MAAM,EAAE,MAAM,CAAC,MAAM,CAAC,EAAE,MAAM,EAAE,YAAY,EAAE,CAAC;aAChD,CAAC,CAAC;YAEH,MAAM,MAAM,GAAG,MAAM,CAAC,MAAM,IAAI,EAAE,CAAC;YACnC,OAAO,UAAU,CAAC,MAAM,EAAE,UAAU,EAAE,CAAC,EAAE,cAAc,CAAC,CAAC;QAC3D,CAAC;KACF,CAAC;AACJ,CAAC;AAED,SAAS,iBAAiB,CAAC,OAA0B;IACnD,MAAM,UAAU,GAAmF,EAAE,CAAC;IACtG,KAAK,MAAM,MAAM,IAAI,OAAO,EAAE,CAAC;QAC7B,MAAM,MAAM,GAAG,MAAM,CAAC,SAAS,EAAE,CAAC;QAClC,KAAK,MAAM,KAAK,IAAI,MAAM,EAAE,CAAC;YAC3B,UAAU,CAAC,IAAI,CAAC;gBACd,QAAQ,EAAE,MAAM,CAAC,EAAE;gBACnB,OAAO,EAAE,KAAK,CAAC,EAAE;gBACjB,IAAI,EAAE,KAAK,CAAC,IAAI;gBAChB,UAAU,EAAE,MAAM,CAAC,IAAI;aACxB,CAAC,CAAC;QACL,CAAC;IACH,CAAC;IACD,OAAO,UAAU,CAAC;AACpB,CAAC;AAED,SAAS,WAAW,CAAC,KAMpB;IACC,MAAM,MAAM,GAAG;QACb,UAAU,KAAK,CAAC,KAAK,EAAE;QACvB,KAAK,CAAC,IAAI,CAAC,CAAC,CAAC,SAAS,KAAK,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC,IAAI;QACzC,cAAc,KAAK,CAAC,IAAI,gCAAgC;QACxD,6CAA6C;KAC9C;SACE,MAAM,CAAC,OAAO,CAAC;SACf,IAAI,CAAC,IAAI,CAAC,CAAC;IAEd,MAAM,MAAM,GAAG,KAAK,CAAC,UAAU;SAC5B,GAAG,CAAC,SAAS,CAAC,EAAE;QACf,MAAM,IAAI,GAAG,SAAS,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,EAAE,KAAK,CAAC,iBAAiB,CAAC,CAAC;QAC9D,OAAO,aAAa,SAAS,CAAC,QAAQ,cAAc,SAAS,CAAC,OAAO,iBAAiB,SAAS,CAAC,UAAU,KAAK,IAAI,EAAE,CAAC;IACxH,CAAC,CAAC;SACD,IAAI,CAAC,aAAa,CAAC,CAAC;IAEvB,OAAO,GAAG,MAAM,gBAAgB,MAAM,EAAE,CAAC;AAC3C,CAAC;AAED,SAAS,UAAU,CACjB,MAAqE,EACrE,UAAwD,EACxD,IAAY,EACZ,cAAuB;IAEvB,MAAM,YAAY,GAAG,IAAI,GAAG,CAAC,UAAU,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,GAAG,CAAC,CAAC,QAAQ,KAAK,CAAC,CAAC,OAAO,EAAE,CAAC,CAAC,CAAC;IACjF,MAAM,IAAI,GAAmB,EAAE,CAAC;IAEhC,KAAK,MAAM,KAAK,IAAI,MAAM,EAAE,CAAC;QAC3B,MAAM,GAAG,GAAG,GAAG,KAAK,CAAC,QAAQ,KAAK,KAAK,CAAC,OAAO,EAAE,CAAC;QAClD,IAAI,CAAC,YAAY,CAAC,GAAG,CAAC,GAAG,CAAC;YAAE,SAAS;QACrC,IAAI,IAAI,CAAC,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,CAAC,QAAQ,KAAK,KAAK,CAAC,QAAQ,IAAI,GAAG,CAAC,OAAO,KAAK,KAAK,CAAC,OAAO,CAAC;YAAE,SAAS;QAEjG,IAAI,CAAC,IAAI,CAAC;YACR,QAAQ,EAAE,KAAK,CAAC,QAAQ;YACxB,OAAO,EAAE,KAAK,CAAC,OAAO;YACtB,IAAI,EAAE,IAAI,CAAC,MAAM,GAAG,CAAC;YACrB,MAAM,EAAE,cAAc,CAAC,CAAC,CAAC,KAAK,CAAC,MAAM,IAAI,EAAE,CAAC,CAAC,CAAC,SAAS;SACxD,CAAC,CAAC;QAEH,IAAI,IAAI,CAAC,MAAM,IAAI,IAAI;YAAE,MAAM;IACjC,CAAC;IAED,OAAO,IAAI,CAAC;AACd,CAAC"}
|
package/dist/source.d.ts
ADDED
|
@@ -0,0 +1,13 @@
|
|
|
1
|
+
import type { Chunker, KnowledgeSource } from './types.js';
|
|
2
|
+
export declare function createStaticKnowledgeSource(input: {
|
|
3
|
+
id: string;
|
|
4
|
+
name: string;
|
|
5
|
+
description?: string;
|
|
6
|
+
content: string;
|
|
7
|
+
chunker?: Chunker;
|
|
8
|
+
chunkOptions?: {
|
|
9
|
+
maxChars?: number;
|
|
10
|
+
overlapChars?: number;
|
|
11
|
+
};
|
|
12
|
+
}): KnowledgeSource;
|
|
13
|
+
//# sourceMappingURL=source.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"source.d.ts","sourceRoot":"","sources":["../src/source.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,OAAO,EAAE,eAAe,EAAkB,MAAM,YAAY,CAAC;AAG3E,wBAAgB,2BAA2B,CAAC,KAAK,EAAE;IACjD,EAAE,EAAE,MAAM,CAAC;IACX,IAAI,EAAE,MAAM,CAAC;IACb,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,OAAO,EAAE,MAAM,CAAC;IAChB,OAAO,CAAC,EAAE,OAAO,CAAC;IAClB,YAAY,CAAC,EAAE;QAAE,QAAQ,CAAC,EAAE,MAAM,CAAC;QAAC,YAAY,CAAC,EAAE,MAAM,CAAA;KAAE,CAAC;CAC7D,GAAG,eAAe,CAelB"}
|
package/dist/source.js
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
1
|
+
import { createMarkdownChunker } from './chunkers.js';
|
|
2
|
+
export function createStaticKnowledgeSource(input) {
|
|
3
|
+
const chunker = input.chunker ?? createMarkdownChunker(input.chunkOptions);
|
|
4
|
+
const chunks = chunker.chunk(input.content, input.chunkOptions);
|
|
5
|
+
return {
|
|
6
|
+
id: input.id,
|
|
7
|
+
name: input.name,
|
|
8
|
+
description: input.description,
|
|
9
|
+
getChunks() {
|
|
10
|
+
return chunks;
|
|
11
|
+
},
|
|
12
|
+
dumpContent() {
|
|
13
|
+
return input.content;
|
|
14
|
+
},
|
|
15
|
+
};
|
|
16
|
+
}
|
|
17
|
+
//# sourceMappingURL=source.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"source.js","sourceRoot":"","sources":["../src/source.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,qBAAqB,EAAE,MAAM,eAAe,CAAC;AAEtD,MAAM,UAAU,2BAA2B,CAAC,KAO3C;IACC,MAAM,OAAO,GAAG,KAAK,CAAC,OAAO,IAAI,qBAAqB,CAAC,KAAK,CAAC,YAAY,CAAC,CAAC;IAC3E,MAAM,MAAM,GAAG,OAAO,CAAC,KAAK,CAAC,KAAK,CAAC,OAAO,EAAE,KAAK,CAAC,YAAY,CAAC,CAAC;IAEhE,OAAO;QACL,EAAE,EAAE,KAAK,CAAC,EAAE;QACZ,IAAI,EAAE,KAAK,CAAC,IAAI;QAChB,WAAW,EAAE,KAAK,CAAC,WAAW;QAC9B,SAAS;YACP,OAAO,MAAM,CAAC;QAChB,CAAC;QACD,WAAW;YACT,OAAO,KAAK,CAAC,OAAO,CAAC;QACvB,CAAC;KACF,CAAC;AACJ,CAAC"}
|
package/dist/types.d.ts
ADDED
|
@@ -0,0 +1,40 @@
|
|
|
1
|
+
import type { LanguageModel } from 'ai';
|
|
2
|
+
export type KnowledgeChunk = {
|
|
3
|
+
id: string;
|
|
4
|
+
text: string;
|
|
5
|
+
meta?: Record<string, unknown>;
|
|
6
|
+
};
|
|
7
|
+
export interface KnowledgeSource {
|
|
8
|
+
id: string;
|
|
9
|
+
name: string;
|
|
10
|
+
description?: string;
|
|
11
|
+
getChunks(): KnowledgeChunk[];
|
|
12
|
+
dumpContent?(): string;
|
|
13
|
+
}
|
|
14
|
+
export interface Chunker {
|
|
15
|
+
chunk(text: string, options?: ChunkOptions): KnowledgeChunk[];
|
|
16
|
+
}
|
|
17
|
+
export type ChunkOptions = {
|
|
18
|
+
maxChars?: number;
|
|
19
|
+
overlapChars?: number;
|
|
20
|
+
};
|
|
21
|
+
export type RetrievalHit = {
|
|
22
|
+
sourceId: string;
|
|
23
|
+
chunkId: string;
|
|
24
|
+
rank: number;
|
|
25
|
+
score?: number;
|
|
26
|
+
reason?: string;
|
|
27
|
+
};
|
|
28
|
+
export interface KnowledgeRetriever {
|
|
29
|
+
retrieve(query: string, sources: KnowledgeSource[], options?: {
|
|
30
|
+
topK?: number;
|
|
31
|
+
hint?: string;
|
|
32
|
+
}): Promise<RetrievalHit[]>;
|
|
33
|
+
}
|
|
34
|
+
export type LLMRetrieverOptions = {
|
|
35
|
+
model: LanguageModel;
|
|
36
|
+
topK?: number;
|
|
37
|
+
includeReasons?: boolean;
|
|
38
|
+
candidateMaxChars?: number;
|
|
39
|
+
};
|
|
40
|
+
//# sourceMappingURL=types.d.ts.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"types.d.ts","sourceRoot":"","sources":["../src/types.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,IAAI,CAAC;AAExC,MAAM,MAAM,cAAc,GAAG;IAC3B,EAAE,EAAE,MAAM,CAAC;IACX,IAAI,EAAE,MAAM,CAAC;IACb,IAAI,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;CAChC,CAAC;AAEF,MAAM,WAAW,eAAe;IAC9B,EAAE,EAAE,MAAM,CAAC;IACX,IAAI,EAAE,MAAM,CAAC;IACb,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,SAAS,IAAI,cAAc,EAAE,CAAC;IAC9B,WAAW,CAAC,IAAI,MAAM,CAAC;CACxB;AAED,MAAM,WAAW,OAAO;IACtB,KAAK,CAAC,IAAI,EAAE,MAAM,EAAE,OAAO,CAAC,EAAE,YAAY,GAAG,cAAc,EAAE,CAAC;CAC/D;AAED,MAAM,MAAM,YAAY,GAAG;IACzB,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,YAAY,CAAC,EAAE,MAAM,CAAC;CACvB,CAAC;AAEF,MAAM,MAAM,YAAY,GAAG;IACzB,QAAQ,EAAE,MAAM,CAAC;IACjB,OAAO,EAAE,MAAM,CAAC;IAChB,IAAI,EAAE,MAAM,CAAC;IACb,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,MAAM,CAAC,EAAE,MAAM,CAAC;CACjB,CAAC;AAEF,MAAM,WAAW,kBAAkB;IACjC,QAAQ,CACN,KAAK,EAAE,MAAM,EACb,OAAO,EAAE,eAAe,EAAE,EAC1B,OAAO,CAAC,EAAE;QAAE,IAAI,CAAC,EAAE,MAAM,CAAC;QAAC,IAAI,CAAC,EAAE,MAAM,CAAA;KAAE,GACzC,OAAO,CAAC,YAAY,EAAE,CAAC,CAAC;CAC5B;AAED,MAAM,MAAM,mBAAmB,GAAG;IAChC,KAAK,EAAE,aAAa,CAAC;IACrB,IAAI,CAAC,EAAE,MAAM,CAAC;IACd,cAAc,CAAC,EAAE,OAAO,CAAC;IACzB,iBAAiB,CAAC,EAAE,MAAM,CAAC;CAC5B,CAAC"}
|
package/dist/types.js
ADDED
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"file":"types.js","sourceRoot":"","sources":["../src/types.ts"],"names":[],"mappings":""}
|
package/package.json
ADDED
|
@@ -0,0 +1,36 @@
|
|
|
1
|
+
{
|
|
2
|
+
"name": "@ariaflowagents/rag",
|
|
3
|
+
"version": "0.4.3",
|
|
4
|
+
"description": "RAG primitives for AriaFlow",
|
|
5
|
+
"type": "module",
|
|
6
|
+
"main": "dist/index.js",
|
|
7
|
+
"types": "dist/index.d.ts",
|
|
8
|
+
"exports": {
|
|
9
|
+
".": {
|
|
10
|
+
"types": "./dist/index.d.ts",
|
|
11
|
+
"default": "./dist/index.js"
|
|
12
|
+
}
|
|
13
|
+
},
|
|
14
|
+
"peerDependencies": {
|
|
15
|
+
"ai": "^6.0.0",
|
|
16
|
+
"zod": "^3.0.0"
|
|
17
|
+
},
|
|
18
|
+
"devDependencies": {
|
|
19
|
+
"@types/node": "^20.11.0",
|
|
20
|
+
"ai": "^6.0.0",
|
|
21
|
+
"typescript": "^5.3.0",
|
|
22
|
+
"zod": "^3.23.0"
|
|
23
|
+
},
|
|
24
|
+
"scripts": {
|
|
25
|
+
"build": "tsc -p tsconfig.json",
|
|
26
|
+
"clean": "rm -rf dist",
|
|
27
|
+
"prepublishOnly": "npm run build"
|
|
28
|
+
},
|
|
29
|
+
"publishConfig": {
|
|
30
|
+
"access": "public"
|
|
31
|
+
},
|
|
32
|
+
"files": [
|
|
33
|
+
"dist",
|
|
34
|
+
"README.md"
|
|
35
|
+
]
|
|
36
|
+
}
|