Spaces:
Sleeping
Sleeping
File size: 19,543 Bytes
4862c84 76309ef 4862c84 76309ef 4862c84 76309ef 4862c84 76309ef 4862c84 76309ef 4862c84 76309ef 4862c84 76309ef 4862c84 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 |
from __future__ import annotations
"""Conversation display helpers for vis_gradio.
This module encapsulates everything related to:
• safely parsing model responses (lists / dicts / JSON strings)
• pretty-printing embedded dictionaries for readability
• converting multiple conversation formats to the OpenAI chat list format
• rendering that list as HTML (including accordion grouping + raw JSON viewer).
Moving this logic out of utils.py keeps the latter lean and focussed on general
analytics utilities.
"""
from typing import List, Dict, Any
import ast
import json
import html
import markdown
import re
__all__: List[str] = [
"convert_to_openai_format",
"display_openai_conversation_html",
"pretty_print_embedded_dicts",
]
# ---------------------------------------------------------------------------
# Pretty-printing helpers
# ---------------------------------------------------------------------------
def _find_balanced_spans(text: str):
"""Return (start, end) spans of balanced {...} or [...] regions in *text*."""
spans, stack = [], []
for i, ch in enumerate(text):
if ch in "{[":
stack.append((ch, i))
elif ch in "]}" and stack:
opener, start = stack.pop()
if (opener, ch) in {("{", "}"), ("[", "]")} and not stack:
spans.append((start, i + 1))
return spans
def _try_parse_slice(slice_: str):
"""Attempt to parse *slice_* into a Python object; return None on failure."""
try:
return ast.literal_eval(slice_)
except Exception:
try:
return json.loads(slice_)
except Exception:
return None
def _find_code_spans(text: str) -> List[tuple]:
"""Return spans for markdown code regions to be preserved as-is.
Includes:
- fenced code blocks delimited by ``` ... ```
- inline code segments delimited by `...`
"""
spans: List[tuple] = []
# Fenced blocks ``` ... ``` (language spec allowed after opening fence)
idx = 0
while True:
start = text.find("```", idx)
if start == -1:
break
# Find the end fence
end = text.find("```", start + 3)
if end == -1:
# Unclosed fence: treat rest of string as code
spans.append((start, len(text)))
break
spans.append((start, end + 3))
idx = end + 3
# Inline code `...`
for m in re.finditer(r"`[^`]*`", text, flags=re.DOTALL):
spans.append((m.start(), m.end()))
# Sort and merge overlapping spans
spans.sort()
merged: List[tuple] = []
for s, e in spans:
if not merged or s > merged[-1][1]:
merged.append((s, e))
else:
merged[-1] = (merged[-1][0], max(merged[-1][1], e))
return merged
def _is_inside_any_span(start: int, end: int, spans: List[tuple]) -> bool:
for s, e in spans:
if start >= s and end <= e:
return True
return False
def pretty_print_embedded_dicts(text: str) -> str:
"""Replace dicts or list-of-dicts with a `<pre>` block, except inside code.
Dict-like regions that fall within markdown code spans (inline backticks
or fenced code blocks) are left untouched so code examples render verbatim.
"""
if not text:
return text
code_spans = _find_code_spans(text)
new_parts, last_idx = [], 0
for start, end in _find_balanced_spans(text):
candidate = text[start:end]
parsed = _try_parse_slice(candidate)
is_good = isinstance(parsed, dict) or (
isinstance(parsed, list) and parsed and all(isinstance(d, dict) for d in parsed)
)
if is_good and not _is_inside_any_span(start, end, code_spans):
new_parts.append(html.escape(text[last_idx:start], quote=False))
pretty = json.dumps(parsed, indent=2, ensure_ascii=False)
new_parts.append(
f"<pre style='background:#f8f9fa;padding:10px;border-radius:4px;overflow-x:auto;'>{pretty}</pre>"
)
last_idx = end
new_parts.append(html.escape(text[last_idx:], quote=False))
return "".join(new_parts)
# ---------------------------------------------------------------------------
# Format conversion
# ---------------------------------------------------------------------------
def convert_to_openai_format(response_data: Any):
"""Convert various response payloads into the OpenAI chat format list."""
if isinstance(response_data, list):
return response_data
if isinstance(response_data, str):
# Try Python literal first (handles single quotes)
try:
parsed = ast.literal_eval(response_data)
if isinstance(parsed, list):
return parsed
except (ValueError, SyntaxError):
pass
# Try JSON
try:
parsed = json.loads(response_data)
if isinstance(parsed, list):
return parsed
except json.JSONDecodeError:
pass
# Fallback plain-text assistant message
return [{"role": "assistant", "content": response_data}]
# Fallback for any other type
return [{"role": "assistant", "content": str(response_data)}]
# ---------------------------------------------------------------------------
# HTML rendering
# ---------------------------------------------------------------------------
def _markdown(text: str, *, pretty_print_dicts: bool = True) -> str:
"""Render markdown, optionally pretty-printing any embedded dicts."""
processed = pretty_print_embedded_dicts(text) if pretty_print_dicts else html.escape(text, quote=False)
# Configure extensions for proper code block handling
extensions = ["fenced_code"]
extension_configs = {}
try:
import pygments
extensions.append("codehilite")
extension_configs['codehilite'] = {
'css_class': 'highlight',
'use_pygments': True,
'guess_lang': True,
'linenums': False
}
except ImportError:
pass
# Convert newlines to <br> only outside of code blocks
# Process fenced code blocks first, then handle line breaks
result = markdown.markdown(processed, extensions=extensions, extension_configs=extension_configs)
# Add line breaks for non-code content (simple approach)
# This replaces single newlines with <br> but preserves code blocks
import re
# Split by code blocks to avoid affecting them
code_block_pattern = r'(<pre[^>]*>.*?</pre>|<code[^>]*>.*?</code>)'
parts = re.split(code_block_pattern, result, flags=re.DOTALL)
for i in range(0, len(parts), 2): # Process non-code parts only
if i < len(parts):
# Replace single newlines with <br>, but not double newlines (paragraphs)
parts[i] = re.sub(r'(?<!\n)\n(?!\n)', '<br>\n', parts[i])
return ''.join(parts)
def display_openai_conversation_html(conversation_data: List[Dict[str, Any]], *, use_accordion: bool = True, pretty_print_dicts: bool = True, evidence: Any = None) -> str:
"""Convert an OpenAI-style conversation list into styled HTML for Gradio."""
from .examples_helpers import annotate_text_with_evidence_placeholders, HIGHLIGHT_START, HIGHLIGHT_END
if not conversation_data:
return "<p>No conversation data available</p>"
# Collapsed raw JSON section for debugging
raw_json = json.dumps(conversation_data, indent=2, ensure_ascii=False)
html_out = f"""
<details style="margin: 8px 0;">
<summary style="cursor: pointer; font-weight: 600;">
Click to see raw response ({len(conversation_data)})
</summary>
<div style="padding: 8px 15px;">
<pre style="white-space: pre-wrap; word-wrap: break-word; background: #f8f9fa; padding: 10px; border-radius: 4px; overflow-x: auto;">{raw_json}</pre>
</div>
</details>
"""
role_colors = {
"system": "#ff6b6b",
"info": "#4ecdc4",
"assistant": "#45b7d1",
"tool": "#96ceb4",
"user": "#feca57",
}
def _maybe_annotate(content_str: str) -> str:
if evidence is None or not isinstance(content_str, str) or not content_str.strip():
return content_str
return annotate_text_with_evidence_placeholders(content_str, evidence)
def _replace_placeholders_with_mark(html_str: str) -> str:
if not html_str:
return html_str
return (
html_str
.replace(HIGHLIGHT_START, "<mark class=\"evidence-highlight\">")
.replace(HIGHLIGHT_END, "</mark>")
)
def _format_msg(role: str, content: Any) -> str:
if isinstance(content, dict) or (isinstance(content, list) and content and all(isinstance(d, dict) for d in content)):
if pretty_print_dicts:
content_html = (
f"<pre style='background: #f8f9fa; padding: 10px; border-radius: 4px; overflow-x: auto;'>{json.dumps(content, indent=2, ensure_ascii=False)}</pre>"
)
else:
content_html = f"<code>{html.escape(json.dumps(content, ensure_ascii=False))}</code>"
elif isinstance(content, str):
# Insert highlight placeholders before markdown so offsets make sense in plain text
annotated = _maybe_annotate(content)
content_html = _markdown(annotated, pretty_print_dicts=pretty_print_dicts)
# Convert placeholders to <mark> after markdown
content_html = _replace_placeholders_with_mark(content_html)
elif content is None:
content_html = "<em>(No content)</em>"
else:
content_html = str(content)
color = role_colors.get(role.lower(), "#95a5a6")
return (
f"<div style='border-left: 4px solid {color}; margin: 8px 0; background-color: #ffffff; padding: 12px; border-radius: 0 8px 8px 0;'>"
f"<div style='font-weight: 600; color: {color}; margin-bottom: 8px; text-transform: capitalize; font-size: 14px;'>{role}</div>"
f"<div style='color: #333; line-height: 1.6; font-family: \"Segoe UI\", Tahoma, Geneva, Verdana, sans-serif;'>{content_html}</div>"
"</div>"
)
if use_accordion:
system_msgs, info_msgs, other_msgs = [], [], []
for m in conversation_data:
if not isinstance(m, dict):
continue
role = m.get("role", "unknown").lower()
content = m.get("content", "")
if isinstance(content, dict) and "text" in content:
content = content["text"]
if role == "system":
system_msgs.append((role, content))
elif role == "info":
info_msgs.append((role, content))
else:
other_msgs.append((role, content))
def _accordion(title: str, items: List):
if not items:
return ""
inner = "".join(_format_msg(r, c) for r, c in items)
return (
f"<details style='margin: 8px 0;'>"
f"<summary style='cursor: pointer; font-weight: 600;'>"
f"{html.escape(title)} ({len(items)})" # e.g. "Click to see system messages (3)"
f"</summary>"
f"<div style='padding: 8px 15px;'>{inner}</div>"
"</details>"
)
html_out += _accordion("Click to see system messages", system_msgs)
html_out += _accordion("Click to see info messages", info_msgs)
for r, c in other_msgs:
html_out += _format_msg(r, c)
else:
# No accordion: just render everything
for m in conversation_data:
if not isinstance(m, dict):
continue
role = m.get("role", "unknown").lower()
content = m.get("content", "")
if isinstance(content, dict) and "text" in content:
content = content["text"]
html_out += _format_msg(role, content)
# CSS for proper code block styling and summary hover effects
css_styles = """
<style>
.evidence-highlight { background: #ffff8b; padding: 0 2px; }
:root {
/* Code block color palette - GitHub Light inspired */
--code-bg: #f6f8fa;
--code-text: #24292f;
--code-comment: #6a737d;
--code-keyword: #d73a49;
--code-string: #032f62;
--code-number: #005cc5;
--code-operator: #24292f;
--code-function: #6f42c1;
--code-border: #d0d7de;
/* Inline code colors - same light theme */
--inline-code-bg: #f3f4f6;
--inline-code-text: #24292f;
--inline-code-border: #d1d5db;
/* Code block structure */
--code-border-radius: 8px;
--code-padding: 16px;
--code-font-size: 14px;
--code-line-height: 1.5;
--code-font-family: 'JetBrains Mono', 'Fira Code', 'Cascadia Code', 'SF Mono', Consolas, 'Liberation Mono', Menlo, Courier, monospace;
}
/* Base code styling */
pre, code {
font-family: var(--code-font-family) !important;
font-size: var(--code-font-size) !important;
line-height: var(--code-line-height) !important;
font-variant-ligatures: normal !important;
-webkit-font-smoothing: antialiased !important;
-moz-osx-font-smoothing: grayscale !important;
}
/* Fenced code blocks - light theme */
.highlight, .codehilite, pre.highlight, pre.codehilite,
.language-python, .language-text, .language-bash {
background: var(--code-bg) !important;
color: var(--code-text) !important;
border: 1px solid var(--code-border) !important;
border-radius: var(--code-border-radius) !important;
padding: var(--code-padding) !important;
margin: 12px 0 !important;
overflow-x: auto !important;
box-shadow: 0 2px 4px rgba(0, 0, 0, 0.05) !important;
position: relative !important;
white-space: pre !important;
display: block !important;
}
.highlight pre, .codehilite pre {
background: transparent !important;
color: inherit !important;
margin: 0 !important;
padding: 0 !important;
border: none !important;
border-radius: 0 !important;
overflow: visible !important;
white-space: pre !important;
display: block !important;
}
/* Ensure code blocks preserve formatting */
.highlight code, .codehilite code {
white-space: pre !important;
display: block !important;
padding: 0 !important;
margin: 0 !important;
background: transparent !important;
border: none !important;
font-size: inherit !important;
line-height: inherit !important;
}
/* Add language label for fenced blocks */
.highlight::before, .codehilite::before {
content: 'python';
position: absolute;
top: 8px;
right: 12px;
background: rgba(0, 0, 0, 0.05);
color: #586069;
padding: 2px 8px;
border-radius: 4px;
font-size: 11px;
font-weight: 500;
text-transform: uppercase;
letter-spacing: 0.5px;
}
/* Syntax highlighting for Python - Light theme */
.highlight .k, .codehilite .k, /* keywords */
.highlight .kn, .codehilite .kn, /* keyword.namespace */
.highlight .kp, .codehilite .kp, /* keyword.pseudo */
.highlight .kr, .codehilite .kr, /* keyword.reserved */
.highlight .kt, .codehilite .kt /* keyword.type */
{
color: var(--code-keyword) !important;
font-weight: 600 !important;
}
.highlight .s, .codehilite .s, /* strings */
.highlight .s1, .codehilite .s1, /* string.single */
.highlight .s2, .codehilite .s2, /* string.double */
.highlight .se, .codehilite .se /* string.escape */
{
color: var(--code-string) !important;
}
.highlight .c, .codehilite .c, /* comments */
.highlight .c1, .codehilite .c1, /* comment.single */
.highlight .cm, .codehilite .cm /* comment.multiline */
{
color: var(--code-comment) !important;
font-style: italic !important;
}
.highlight .m, .codehilite .m, /* numbers */
.highlight .mi, .codehilite .mi, /* number.integer */
.highlight .mf, .codehilite .mf, /* number.float */
.highlight .mo, .codehilite .mo /* number.octal */
{
color: var(--code-number) !important;
font-weight: 600 !important;
}
.highlight .nf, .codehilite .nf, /* function names */
.highlight .fm, .codehilite .fm /* function.magic */
{
color: var(--code-function) !important;
font-weight: 600 !important;
}
.highlight .o, .codehilite .o, /* operators */
.highlight .ow, .codehilite .ow /* operator.word */
{
color: var(--code-operator) !important;
}
/* Inline code - light theme */
p code, li code, div code, span code,
h1 code, h2 code, h3 code, h4 code, h5 code, h6 code {
background: var(--inline-code-bg) !important;
color: var(--inline-code-text) !important;
border: 1px solid var(--inline-code-border) !important;
padding: 2px 6px !important;
border-radius: 4px !important;
font-size: 0.9em !important;
font-weight: 600 !important;
white-space: nowrap !important;
box-shadow: none !important;
display: inline !important;
}
/* Code blocks inside paragraphs should not be treated as inline */
p pre, li pre, div pre {
background: var(--code-bg) !important;
color: var(--code-text) !important;
border: 1px solid var(--code-border) !important;
border-radius: var(--code-border-radius) !important;
padding: var(--code-padding) !important;
margin: 8px 0 !important;
white-space: pre !important;
overflow-x: auto !important;
display: block !important;
}
/* Scrollbar styling for code blocks - light theme */
.highlight::-webkit-scrollbar, .codehilite::-webkit-scrollbar,
pre::-webkit-scrollbar {
height: 8px !important;
background: #f1f3f4 !important;
border-radius: 4px !important;
}
.highlight::-webkit-scrollbar-thumb, .codehilite::-webkit-scrollbar-thumb,
pre::-webkit-scrollbar-thumb {
background: #c1c8cd !important;
border-radius: 4px !important;
}
.highlight::-webkit-scrollbar-thumb:hover, .codehilite::-webkit-scrollbar-thumb:hover,
pre::-webkit-scrollbar-thumb:hover {
background: #a8b3ba !important;
}
</style>
"""
if use_accordion:
css_styles += """
/* Accordion styling */
details > summary {
list-style: none !important;
cursor: pointer !important;
}
details > summary:hover {
background-color: transparent !important;
box-shadow: none !important;
transform: none !important;
}
details > summary::-webkit-details-marker,
details > summary::marker {
display: none !important;
}
"""
css_styles += "</style>"
html_out = css_styles + html_out
return html_out |