File size: 5,445 Bytes
cedc2e4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d2509c7
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
import os
import json
import pickle
from typing import List, Dict, Tuple

import numpy as np
import faiss
from sentence_transformers import SentenceTransformer
from pypdf import PdfReader

from .config import (
    DATA_DIR,
    URLS_PATH,
    FAISS_INDEX_PATH,
    DOCSTORE_PATH,
    EMBED_MODEL_NAME,
)
from .fetcher import fetch_page_text


DOCS_DIR = os.path.join(DATA_DIR, "docs")


def ensure_data_dir():
    os.makedirs(DATA_DIR, exist_ok=True)
    os.makedirs(DOCS_DIR, exist_ok=True)  # safe even if empty


def load_urls() -> List[str]:
    """
    Expects data/urls.json like:
    { "urls": ["https://...", "https://..."] }
    """
    if not os.path.exists(URLS_PATH):
        # If urls.json missing, we allow ingestion to continue with local docs only
        return []

    with open(URLS_PATH, "r", encoding="utf-8") as f:
        obj = json.load(f)

    urls = obj.get("urls", [])
    return [u.strip() for u in urls if isinstance(u, str) and u.strip()]


def chunk_text(text: str, chunk_size_words: int = 900, overlap_words: int = 150) -> List[str]:
    """
    Simple word-based chunking (fast + reliable).
    """
    text = (text or "").strip()
    if not text:
        return []

    words = text.split()
    chunks = []
    i = 0
    step = max(1, chunk_size_words - overlap_words)

    while i < len(words):
        chunk = words[i:i + chunk_size_words]
        chunks.append(" ".join(chunk))
        i += step

    return chunks


# -------------------------
# URL ingestion
# -------------------------
def build_docs_from_urls(urls: List[str]) -> List[Dict]:
    docs: List[Dict] = []
    for url in urls:
        try:
            page = fetch_page_text(url, use_cache=True)
            chunks = chunk_text(page.get("text", ""))

            for idx, ch in enumerate(chunks):
                docs.append({
                    "text": ch,
                    "meta": {
                        "source_type": "url",
                        "url": page.get("url", url),
                        "title": page.get("title", url),
                        "chunk": idx,
                    }
                })
        except Exception:
            # skip bad URLs but continue ingestion
            continue
    return docs


# -------------------------
# Local docs ingestion
# -------------------------
def list_local_files() -> List[str]:
    """
    Reads local files from data/docs/
    Supported: .txt, .md, .pdf (text-based PDFs)
    """
    if not os.path.exists(DOCS_DIR):
        return []

    paths = []
    for name in os.listdir(DOCS_DIR):
        p = os.path.join(DOCS_DIR, name)
        if not os.path.isfile(p):
            continue
        ext = os.path.splitext(name)[1].lower()
        if ext in [".txt", ".md", ".pdf"]:
            paths.append(p)
    return sorted(paths)


def read_text_file(path: str) -> str:
    with open(path, "r", encoding="utf-8", errors="ignore") as f:
        return f.read()


def read_pdf_text(path: str) -> str:
    """
    Works best on selectable-text PDFs.
    Scanned/image-only PDFs will extract very little.
    """
    reader = PdfReader(path)
    parts = []
    for page in reader.pages:
        try:
            parts.append(page.extract_text() or "")
        except Exception:
            continue
    return "\n".join(parts).strip()


def build_docs_from_files(file_paths: List[str]) -> List[Dict]:
    docs: List[Dict] = []

    for path in file_paths:
        name = os.path.basename(path)
        ext = os.path.splitext(name)[1].lower()

        try:
            if ext in [".txt", ".md"]:
                text = read_text_file(path)
            elif ext == ".pdf":
                text = read_pdf_text(path)
            else:
                continue
        except Exception:
            continue

        chunks = chunk_text(text)
        for idx, ch in enumerate(chunks):
            docs.append({
                "text": ch,
                "meta": {
                    "source_type": "file",
                    "url": f"file://{name}",
                    "title": name,
                    "chunk": idx,
                }
            })

    return docs


# -------------------------
# Index building
# -------------------------
def build_faiss_index(docs: List[Dict]) -> None:
    model = SentenceTransformer(EMBED_MODEL_NAME)

    texts = [d["text"] for d in docs]
    emb = model.encode(texts, normalize_embeddings=True, show_progress_bar=True)
    emb = np.array(emb, dtype="float32")

    index = faiss.IndexFlatIP(emb.shape[1])
    index.add(emb)

    faiss.write_index(index, FAISS_INDEX_PATH)

    with open(DOCSTORE_PATH, "wb") as f:
        pickle.dump(docs, f)


def run_ingestion():
    ensure_data_dir()

    urls = load_urls()
    url_docs = build_docs_from_urls(urls) if urls else []

    file_paths = list_local_files()
    file_docs = build_docs_from_files(file_paths) if file_paths else []

    docs = url_docs + file_docs

    if not docs:
        raise RuntimeError(
            "No documents found.\n"
            "- Add URLs to data/urls.json OR\n"
            "- Add files to data/docs/ (.txt, .md, .pdf)"
        )

    build_faiss_index(docs)

    print("✅ Ingestion complete")
    print(f"URLs: {len(urls)}")
    print(f"Local files: {len(file_paths)}")
    print(f"Chunks: {len(docs)}")
    print(f"Saved index: {FAISS_INDEX_PATH}")
    print(f"Saved docs:  {DOCSTORE_PATH}")


if __name__ == "__main__":
    run_ingestion()