Fix export issues (#1018)

lumen/ai/export.py

@@ -1,13 +1,17 @@
+import base64
 import datetime as dt
 import json
+import os
 
+from io import BytesIO
 from textwrap import dedent
 from typing import Any
 
 import nbformat
 
 from panel import Column
 from panel.chat import ChatMessage, ChatStep
+from panel.pane.image import ImageBase
 
 from lumen.ai.views import LumenOutput
 from lumen.config import config
@@ -35,12 +39,41 @@ def make_preamble(preamble: str, extensions: list[str]):
     imports = nbformat.v4.new_code_cell(source=source)
     return [header, imports]
 
-def format_markdown(msg: ChatMessage):
-    if msg.avatar.startswith('https://'):
-        avatar = f'<img src="{msg.avatar}" width=45 height=45></img>'
+
+def serialize_avatar(avatar: str | BytesIO, size: int = 45) -> str:
+    """
+    Process different types of avatar inputs into HTML img tag or text span.
+
+    Args:
+        avatar: Avatar source (URL, BytesIO, PIL Image, or text)
+        size: Desired width and height of the avatar in pixels
+
+    Returns:
+        HTML string representing the avatar
+    """
+    if isinstance(avatar, ImageBase):
+        avatar = avatar.object
+    if isinstance(avatar, BytesIO):
+        avatar = avatar.getvalue()
+
+    if isinstance(avatar, str):
+        if avatar.startswith(('http://', 'https://')):
+            return f'<img src="{avatar}" width={size} height={size} alt="avatar" style="border-radius: 50%;"></img>'
+        elif os.path.exists(avatar):
+            avatar_path = os.path.abspath(avatar)
+            return f'<img src="{avatar_path}" width={size} height={size} alt="avatar" style="border-radius: 50%;"></img>'
+        else:
+            return f'<span class="text-avatar" style="width: {size}px; height: {size}px; display: inline-flex; align-items: center; justify-content: center; background-color: #f0f0f0; border-radius: 50%;">{avatar[:2].upper()}</span>'
+    elif isinstance(avatar, bytes):
+        img_data = base64.b64encode(avatar).decode()
+        return f'<img src="data:image/png;base64,{img_data}" width={size} height={size} alt="avatar" style="border-radius: 50%;"></img>'
     else:
-        avatar = f'<span>{msg.avatar}</span>'
-    header = f'<div style="display: flex; flex-direction: row; font-weight: bold; font-size: 2em;">{avatar}<span style="margin-left: 0.5em">{msg.user}</span></div>'
+        return f'<span class="text-avatar" style="width: {size}px; height: {size}px; display: inline-flex; align-items: center; justify-content: center; background-color: #f0f0f0; border-radius: 50%;">{str(avatar)[:2].upper()}</span>'
+
+
+def format_markdown(msg: ChatMessage):
+    avatar_html = serialize_avatar(msg.avatar)
+    header = f'<div style="display: flex; flex-direction: row; font-weight: bold; font-size: 2em;">{avatar_html}<span style="margin-left: 0.5em">{msg.user}</span></div>'
     prefix = '\n' if msg.user == 'User' else '\n> '
     content = prefix.join(msg.serialize().split('\n'))
     return [nbformat.v4.new_markdown_cell(source=f'{header}\n{prefix}{content}')]
@@ -67,7 +100,7 @@ def format_output(msg: ChatMessage):
 def render_cells(messages: list[ChatMessage]) -> tuple[Any, list[str]]:
     cells, extensions = [], []
     for msg in messages:
-        if msg.user == 'Help':
+        if msg.user in ("Help", " "):
             continue
         elif isinstance(msg.object, str):
             cells += format_markdown(msg)

lumen/ai/ui.py

@@ -100,7 +100,7 @@ class UI(Viewer):
     llm = param.ClassSelector(class_=Llm, default=OpenAI(), doc="""
         The LLM provider to be used by default""")
 
-    log_level = param.ObjectSelector(default='DEBUG', objects=['DEBUG', 'INFO', 'WARNING', 'ERROR'], doc="""
+    log_level = param.ObjectSelector(default='INFO', objects=['DEBUG', 'INFO', 'WARNING', 'ERROR'], doc="""
         The log level to use.""")
 
     logs_db_path = param.String(default=None, doc="""

lumen/sources/duckdb.py

@@ -126,7 +126,10 @@ def _serialize_tables(self):
             tables[t] = serializer.serialize(tdf)
         return tables
 
-    def _process_sql_paths(self, sql_expr: str) -> str:
+    def _process_sql_paths(self, sql_expr: dict | str) -> str:
+        if isinstance(sql_expr, dict):
+            return {self._process_sql_paths(k): v for k, v in sql_expr.items()}
+
         # Look for read_* patterns like read_parquet, read_csv etc.
         matches = re.finditer(r"read_\w+\('([^']+)'\)", sql_expr)
         processed_sql = sql_expr
@@ -209,6 +212,7 @@ def from_spec(cls, spec: dict[str, Any] | str) -> Source:
             spec['tables'] = {}
         else:
             ephemeral_tables = {}
+
         source = super().from_spec(spec)
         if not ephemeral_tables:
             return source