Spaces:
Sleeping
Sleeping
File size: 11,191 Bytes
d82f721 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 | """Chunkers for the UI GreenMetric RAG system.
Converts raw data sources (markdown, CSV tables) into unified
{content, metadata} dicts ready for embedding and ChromaDB storage.
"""
import pandas as pd
import re
from collections.abc import Callable
# ---------------------------------------------------------------------------
# Generic CSV chunker
# ---------------------------------------------------------------------------
def chunk_csv(
df: pd.DataFrame,
group_col: str,
*,
source: str,
chunk_type: str,
format_fn: Callable[[pd.DataFrame], str],
metadata_fn: Callable[[pd.DataFrame], dict],
) -> list[dict]:
"""Generic group-based CSV chunker.
Groups *df* by *group_col*, then delegates text formatting and
metadata extraction to callables. *source* and *chunk_type* are
injected into the metadata dict automatically.
Parameters:
df: Pre-processed DataFrame (encoding fixes, character
replacements already applied by the caller).
group_col: Column to group by (e.g. ``'no'``, ``'country'``,
``'scope'``).
source: Source identifier injected into every chunk's metadata
(e.g. ``'csv_appendix1'``, ``'csv_table4'``).
chunk_type: Chunk type injected into every chunk's metadata
(e.g. ``'question'``, ``'category'``, ``'reference'``).
format_fn: ``Callable[[pd.DataFrame], str]``
Receives one group at a time. Returns the chunk text.
metadata_fn: ``Callable[[pd.DataFrame], dict]``
Receives one group at a time. Returns domain-specific
metadata keys (``category``, ``question_no``, ...).
``source`` and ``chunk_type`` are added by this
function automatically.
Returns:
list[dict]: One chunk per group. Each chunk has keys ``"content"``
(str) and ``"metadata"`` (dict).
"""
chunks: list[dict] = []
for no, group in df.groupby(group_col, sort=False):
chunks.append({
"content": format_fn(group),
"metadata": metadata_fn(group)
| {"source": source, "chunk_type": chunk_type},
})
return chunks
# ---------------------------------------------------------------------------
# Per-source format / metadata helpers
# ---------------------------------------------------------------------------
# --- appendix1 ----------------------------------------------------------------
def _fmt_appendix1(group: pd.DataFrame) -> str:
no = group.iloc[0]["no"]
text_content = f"""Question {no} — {group.iloc[0]['criteria']}
Category: {group.iloc[0]['category']}
Evidence Required: {group.iloc[0]['evidence_required']}
"""
indicator_code = group.iloc[0]["indicator_code"] if not pd.isna(group.iloc[0]["indicator_code"]) else "Not Available"
max_score = group.iloc[0]["max_score"] if not pd.isna(group.iloc[0]["max_score"]) else "Not Available"
colored = group.iloc[0]["colored"] if not pd.isna(group.iloc[0]["colored"]) else "Not Available"
text_content += f"Indicator Code: {indicator_code}\n"
text_content += f"Max Score: {max_score}\n"
text_content += f"Colored: {colored}\n"
text_content += "Options:\n"
for options in group.itertuples():
if not pd.isna(options.calculated_score):
text_content += f"{options.answer} (Calculated score: {options.calculated_score})\n"
else:
text_content += f"{options.answer} (Calculated score: Not Available)\n"
return text_content
def _meta_appendix1(group: pd.DataFrame) -> dict:
max_score = float(group.iloc[0]["max_score"]) if not pd.isna(group.iloc[0]["max_score"]) else -1.0 # max_score uses -1.0 as sentinel for unscored/ungraded criteria (no real score is negative)
colored = group.iloc[0]["colored"] if not pd.isna(group.iloc[0]["colored"]) else "Not Available"
return {
"category": group.iloc[0]["category"],
"question_no": group.iloc[0]["no"],
"evidence_required": group.iloc[0]["evidence_required"],
"max_score": max_score,
"colored": colored,
}
# --- appendix2 ----------------------------------------------------------------
def _fmt_appendix2(group: pd.DataFrame) -> str:
text_content = f"Category: {group.iloc[0]['element_category']}\n"
text_content += "Existing building category:\n"
for sub_categories, element in group.loc[:, ["gbi_non-residential_existing_building_category", "gbi_non-residential_existing_building_element"]].itertuples(index=False):
if pd.isna(sub_categories) and pd.isna(element):
continue
sub_categories = sub_categories if not pd.isna(sub_categories) else "Not Available"
element = element if not pd.isna(element) else "Not Available"
text_content += f"{sub_categories} | {element}\n"
text_content += "\nNew construction category:\n"
for sub_categories, element in group.loc[:, ["gbi_non-residential_new_construction_(nrnc)_category", "gbi_non-residential_new_construction_(nrnc)_element"]].itertuples(index=False):
if pd.isna(sub_categories) and pd.isna(element):
continue
sub_categories = sub_categories if not pd.isna(sub_categories) else "Not Available"
element = element if not pd.isna(element) else "Not Available"
text_content += f"{sub_categories} | {element}\n"
return text_content
def _meta_appendix2(group: pd.DataFrame) -> dict:
return {
"element_category": group.iloc[0]["element_category"],
}
# --- appendix3 ----------------------------------------------------------------
def _fmt_appendix3(group: pd.DataFrame) -> str:
text_content = f"""
Field code: {group.iloc[0]['field_code']}
Field category: {group.iloc[0]['field_name']}
"""
text_content += "Requirements:\n"
for code, name, description in group.loc[:, ["requirement_code", "requirement_name", "description"]].itertuples(index=False):
text_content += f"{code} | {name}: {description}\n"
return text_content
def _meta_appendix3(group: pd.DataFrame) -> dict:
return {
"field_code": group.iloc[0]["field_code"],
"field_name": group.iloc[0]["field_name"],
}
# --- table1 -------------------------------------------------------------------
def _fmt_table1(group: pd.DataFrame) -> str:
text_content = f"Country: {group.iloc[0]['country']}\n"
text_content += "Universities:\n"
for university in group["university"]:
text_content += f"{university}\n"
return text_content
def _meta_table1(group: pd.DataFrame) -> dict:
return {
"country": group.iloc[0]["country"],
}
# --- table2 -------------------------------------------------------------------
def _fmt_table2(group: pd.DataFrame) -> str:
text_content = f"""
Category: {group.iloc[0]['category']}
Weight(%): {group.iloc[0]['percentage_of_total_points_(%)']}
"""
return text_content
def _meta_table2(group: pd.DataFrame) -> dict:
return {
"category": group.iloc[0]["category"],
}
# --- table4 -------------------------------------------------------------------
def _fmt_table4(group: pd.DataFrame) -> str:
text_content = f"Scope category: {group.iloc[0]['scope']}\n"
text_content += "Emission source:\n"
for source, desc in group.loc[:, ["emission_source", "description_or_examples"]].itertuples(index=False):
text_content += f"{source}: {desc}\n"
return text_content
def _meta_table4(group: pd.DataFrame) -> dict:
return {
"scope": group.iloc[0]["scope"],
}
# ---------------------------------------------------------------------------
# Markdown (PDF) chunker
# ---------------------------------------------------------------------------
def chunk_markdown(filepath: str) -> list[dict]:
"""Split UI GreenMetric guidelines markdown into hierarchical chunks.
Strategy: heading-level structural chunking.
- ``##`` → chunk_type ``"intro"``, category = None
- ``###`` → chunk_type ``"category"``, category from heading text.
Sub-sections (``### a.``, ``### b.``) are appended to
the current chunk rather than split.
- ``####`` → chunk_type ``"question"``, single indicator description.
- ``#####`` → appended to the current chunk, never triggers a split.
A chunk is finalised when the NEXT heading of equal or higher rank is
encountered. Trailing content after the last heading is also captured
as the final chunk.
Returns:
list[dict]: Each chunk has keys ``"content"`` (str) and
``"metadata"`` (dict with ``"source"``, ``"chunk_type"``,
``"category"``).
"""
with open(filepath, "r", encoding="utf-8") as file:
markdown_file = file.read()
current_content = []
chunks = []
current_type = None
current_category = None
for line in markdown_file.splitlines():
if line.startswith("## "):
if current_type is not None:
chunks.append({
"content": "\n".join(current_content).strip(),
"metadata": {
"source": "pdf",
"chunk_type": current_type,
"category": current_category,
},
})
current_category = None
current_type = "intro"
current_content = [line]
elif line.startswith("### "):
if re.match(r"### [a-z]\.", line): # Check if it's not a category of questionnaire (starts with lowercase after ###)
current_content.append(line)
else:
if current_type is not None:
chunks.append({
"content": "\n".join(current_content).strip(),
"metadata": {
"source": "pdf",
"chunk_type": current_type,
"category": current_category,
},
})
current_type = "category"
current_category = line[4:].strip()
current_content = [line]
elif line.startswith("#### "):
if current_type is not None:
chunks.append({
"content": "\n".join(current_content).strip(),
"metadata": {
"source": "pdf",
"chunk_type": current_type,
"category": current_category,
},
})
current_type = "question"
current_content = [line]
elif line.startswith("##### "):
current_content.append(line)
else:
current_content.append(line)
chunks.append({
"content": "\n".join(current_content).strip(),
"metadata": {
"source": "pdf",
"chunk_type": current_type,
"category": current_category,
},
})
return chunks
|