khoj 2.0.0b12.dev5__py3-none-any.whl → 2.0.0b13__py3-none-any.whl

khoj/processor/content/org_mode/org_to_entries.py

@@ -8,7 +8,6 @@ from khoj.database.models import KhojUser
 from khoj.processor.content.org_mode import orgnode
 from khoj.processor.content.org_mode.orgnode import Orgnode
 from khoj.processor.content.text_to_entries import TextToEntries
-from khoj.utils import state
 from khoj.utils.helpers import timer
 from khoj.utils.rawconfig import Entry
 
@@ -103,7 +102,7 @@ class OrgToEntries(TextToEntries):
         # If content is small or content has no children headings, save it as a single entry
         # Note: This is the terminating condition for this recursive function
         if len(TextToEntries.tokenizer(org_content_with_ancestry)) <= max_tokens or not re.search(
-            rf"^\*{{{len(ancestry)+1},}}\s", org_content, re.MULTILINE
+            rf"^\*{{{len(ancestry) + 1},}}\s", org_content, re.MULTILINE
         ):
             orgnode_content_with_ancestry = orgnode.makelist(
                 org_content_with_ancestry, org_file, start_line=start_line, ancestry_lines=len(ancestry)
@@ -195,7 +194,7 @@ class OrgToEntries(TextToEntries):
                 if not entry_heading and parsed_entry.level > 0:
                     base_level = parsed_entry.level
                 # Indent entry by 1 heading level as ancestry is prepended as top level heading
-                heading = f"{'*' * (parsed_entry.level-base_level+2)} {todo_str}" if parsed_entry.level > 0 else ""
+                heading = f"{'*' * (parsed_entry.level - base_level + 2)} {todo_str}" if parsed_entry.level > 0 else ""
                 if parsed_entry.heading:
                     heading += f"{parsed_entry.heading}."
 
@@ -212,10 +211,10 @@ class OrgToEntries(TextToEntries):
                     compiled += f"\t {tags_str}."
 
                 if parsed_entry.closed:
-                    compiled += f'\n Closed on {parsed_entry.closed.strftime("%Y-%m-%d")}.'
+                    compiled += f"\n Closed on {parsed_entry.closed.strftime('%Y-%m-%d')}."
 
                 if parsed_entry.scheduled:
-                    compiled += f'\n Scheduled for {parsed_entry.scheduled.strftime("%Y-%m-%d")}.'
+                    compiled += f"\n Scheduled for {parsed_entry.scheduled.strftime('%Y-%m-%d')}."
 
                 if parsed_entry.hasBody:
                     compiled += f"\n {parsed_entry.body}"

khoj/processor/content/org_mode/orgnode.py

@@ -65,7 +65,7 @@ def makelist(file, filename, start_line: int = 1, ancestry_lines: int = 0) -> Li
     """
     ctr = 0
 
-    if type(file) == str:
+    if isinstance(file, str):
         f = file.splitlines()
     else:
         f = file
@@ -512,11 +512,11 @@ class Orgnode(object):
         if self._closed or self._scheduled or self._deadline:
             n = n + indent
         if self._closed:
-            n = n + f'CLOSED: [{self._closed.strftime("%Y-%m-%d %a")}] '
+            n = n + f"CLOSED: [{self._closed.strftime('%Y-%m-%d %a')}] "
         if self._scheduled:
-            n = n + f'SCHEDULED: <{self._scheduled.strftime("%Y-%m-%d %a")}> '
+            n = n + f"SCHEDULED: <{self._scheduled.strftime('%Y-%m-%d %a')}> "
         if self._deadline:
-            n = n + f'DEADLINE: <{self._deadline.strftime("%Y-%m-%d %a")}> '
+            n = n + f"DEADLINE: <{self._deadline.strftime('%Y-%m-%d %a')}> "
         if self._closed or self._scheduled or self._deadline:
             n = n + "\n"
 

khoj/processor/content/plaintext/plaintext_to_entries.py

@@ -1,6 +1,5 @@
 import logging
 import re
-from pathlib import Path
 from typing import Dict, List, Tuple
 
 import urllib3
@@ -97,7 +96,7 @@ class PlaintextToEntries(TextToEntries):
         for parsed_entry in parsed_entries:
             raw_filename = entry_to_file_map[parsed_entry]
             # Check if raw_filename is a URL. If so, save it as is. If not, convert it to a Path.
-            if type(raw_filename) == str and re.search(r"^https?://", raw_filename):
+            if isinstance(raw_filename, str) and re.search(r"^https?://", raw_filename):
                 # Escape the URL to avoid issues with special characters
                 entry_filename = urllib3.util.parse_url(raw_filename).url
             else:

khoj/processor/content/text_to_entries.py

@@ -27,12 +27,10 @@ logger = logging.getLogger(__name__)
 class TextToEntries(ABC):
     def __init__(self, config: Any = None):
         self.embeddings_model = state.embeddings_model
-        self.config = config
         self.date_filter = DateFilter()
 
     @abstractmethod
-    def process(self, files: dict[str, str], user: KhojUser, regenerate: bool = False) -> Tuple[int, int]:
-        ...
+    def process(self, files: dict[str, str], user: KhojUser, regenerate: bool = False) -> Tuple[int, int]: ...
 
     @staticmethod
     def hash_func(key: str) -> Callable:

khoj/processor/conversation/google/utils.py

@@ -194,7 +194,7 @@ def gemini_completion_with_backoff(
             or not response.candidates[0].content
             or response.candidates[0].content.parts is None
         ):
-            raise ValueError(f"Failed to get response from model.")
+            raise ValueError("Failed to get response from model.")
         raw_content = [part.model_dump() for part in response.candidates[0].content.parts]
         if response.function_calls:
             function_calls = [
@@ -212,7 +212,7 @@ def gemini_completion_with_backoff(
         response = None
         # Handle 429 rate limit errors directly
         if e.code == 429:
-            response_text = f"My brain is exhausted. Can you please try again in a bit?"
+            response_text = "My brain is exhausted. Can you please try again in a bit?"
             # Log the full error details for debugging
             logger.error(f"Gemini ClientError: {e.code} {e.status}. Details: {e.details}")
         # Handle other errors
@@ -361,7 +361,7 @@ def handle_gemini_response(
 
     # Ensure we have a proper list of candidates
     if not isinstance(candidates, list):
-        message = f"\nUnexpected response format. Try again."
+        message = "\nUnexpected response format. Try again."
         stopped = True
         return message, stopped
 

khoj/processor/conversation/openai/gpt.py

@@ -9,6 +9,9 @@ from khoj.processor.conversation.openai.utils import (
     clean_response_schema,
     completion_with_backoff,
     get_structured_output_support,
+    is_openai_api,
+    responses_chat_completion_with_backoff,
+    responses_completion_with_backoff,
     to_openai_tools,
 )
 from khoj.processor.conversation.utils import (
@@ -43,31 +46,52 @@ def send_message_to_model(
     model_kwargs: Dict[str, Any] = {}
     json_support = get_structured_output_support(model, api_base_url)
     if tools and json_support == StructuredOutputSupport.TOOL:
-        model_kwargs["tools"] = to_openai_tools(tools)
+        model_kwargs["tools"] = to_openai_tools(tools, use_responses_api=is_openai_api(api_base_url))
     elif response_schema and json_support >= StructuredOutputSupport.SCHEMA:
         # Drop unsupported fields from schema passed to OpenAI APi
         cleaned_response_schema = clean_response_schema(response_schema)
-        model_kwargs["response_format"] = {
-            "type": "json_schema",
-            "json_schema": {
-                "schema": cleaned_response_schema,
-                "name": response_schema.__name__,
-                "strict": True,
-            },
-        }
+        if is_openai_api(api_base_url):
+            model_kwargs["text"] = {
+                "format": {
+                    "type": "json_schema",
+                    "strict": True,
+                    "name": response_schema.__name__,
+                    "schema": cleaned_response_schema,
+                }
+            }
+        else:
+            model_kwargs["response_format"] = {
+                "type": "json_schema",
+                "json_schema": {
+                    "schema": cleaned_response_schema,
+                    "name": response_schema.__name__,
+                    "strict": True,
+                },
+            }
     elif response_type == "json_object" and json_support == StructuredOutputSupport.OBJECT:
         model_kwargs["response_format"] = {"type": response_type}
 
     # Get Response from GPT
-    return completion_with_backoff(
-        messages=messages,
-        model_name=model,
-        openai_api_key=api_key,
-        api_base_url=api_base_url,
-        deepthought=deepthought,
-        model_kwargs=model_kwargs,
-        tracer=tracer,
-    )
+    if is_openai_api(api_base_url):
+        return responses_completion_with_backoff(
+            messages=messages,
+            model_name=model,
+            openai_api_key=api_key,
+            api_base_url=api_base_url,
+            deepthought=deepthought,
+            model_kwargs=model_kwargs,
+            tracer=tracer,
+        )
+    else:
+        return completion_with_backoff(
+            messages=messages,
+            model_name=model,
+            openai_api_key=api_key,
+            api_base_url=api_base_url,
+            deepthought=deepthought,
+            model_kwargs=model_kwargs,
+            tracer=tracer,
+        )
 
 
 async def converse_openai(
@@ -163,13 +187,26 @@ async def converse_openai(
     logger.debug(f"Conversation Context for GPT: {messages_to_print(messages)}")
 
     # Get Response from GPT
-    async for chunk in chat_completion_with_backoff(
-        messages=messages,
-        model_name=model,
-        temperature=temperature,
-        openai_api_key=api_key,
-        api_base_url=api_base_url,
-        deepthought=deepthought,
-        tracer=tracer,
-    ):
-        yield chunk
+    if is_openai_api(api_base_url):
+        async for chunk in responses_chat_completion_with_backoff(
+            messages=messages,
+            model_name=model,
+            temperature=temperature,
+            openai_api_key=api_key,
+            api_base_url=api_base_url,
+            deepthought=deepthought,
+            tracer=tracer,
+        ):
+            yield chunk
+    else:
+        # For non-OpenAI APIs, use the chat completion method
+        async for chunk in chat_completion_with_backoff(
+            messages=messages,
+            model_name=model,
+            temperature=temperature,
+            openai_api_key=api_key,
+            api_base_url=api_base_url,
+            deepthought=deepthought,
+            tracer=tracer,
+        ):
+            yield chunk