khoj 1.28.3__py3-none-any.whl → 1.28.4.dev92__py3-none-any.whl

khoj/routers/research.py

@@ -11,6 +11,7 @@ from khoj.processor.conversation import prompts
 from khoj.processor.conversation.utils import (
     InformationCollectionIteration,
     clean_json,
+    construct_chat_history,
     construct_iteration_history,
     construct_tool_chat_history,
 )
@@ -19,8 +20,6 @@ from khoj.processor.tools.run_code import run_code
 from khoj.routers.api import extract_references_and_questions
 from khoj.routers.helpers import (
     ChatEvent,
-    construct_chat_history,
-    extract_relevant_info,
     generate_summary_from_files,
     send_message_to_model_wrapper,
 )
@@ -43,38 +42,36 @@ async def apick_next_tool(
     location: LocationData = None,
     user_name: str = None,
     agent: Agent = None,
-    previous_iterations_history: str = None,
+    previous_iterations: List[InformationCollectionIteration] = [],
     max_iterations: int = 5,
     send_status_func: Optional[Callable] = None,
     tracer: dict = {},
+    query_files: str = None,
 ):
-    """
-    Given a query, determine which of the available tools the agent should use in order to answer appropriately. One at a time, and it's able to use subsequent iterations to refine the answer.
-    """
+    """Given a query, determine which of the available tools the agent should use in order to answer appropriately."""
 
+    # Construct tool options for the agent to choose from
     tool_options = dict()
     tool_options_str = ""
-
     agent_tools = agent.input_tools if agent else []
-
     for tool, description in function_calling_description_for_llm.items():
         tool_options[tool.value] = description
         if len(agent_tools) == 0 or tool.value in agent_tools:
             tool_options_str += f'- "{tool.value}": "{description}"\n'
 
+    # Construct chat history with user and iteration history with researcher agent for context
     chat_history = construct_chat_history(conversation_history, agent_name=agent.name if agent else "Khoj")
+    previous_iterations_history = construct_iteration_history(previous_iterations, prompts.previous_iteration)
 
     if query_images:
         query = f"[placeholder for user attached images]\n{query}"
 
+    today = datetime.today()
+    location_data = f"{location}" if location else "Unknown"
     personality_context = (
         prompts.personality_context.format(personality=agent.personality) if agent and agent.personality else ""
     )
 
-    # Extract Past User Message and Inferred Questions from Conversation Log
-    today = datetime.today()
-    location_data = f"{location}" if location else "Unknown"
-
     function_planning_prompt = prompts.plan_function_execution.format(
         tools=tool_options_str,
         chat_history=chat_history,
@@ -87,15 +84,25 @@ async def apick_next_tool(
         max_iterations=max_iterations,
     )
 
-    with timer("Chat actor: Infer information sources to refer", logger):
-        response = await send_message_to_model_wrapper(
-            query=query,
-            context=function_planning_prompt,
-            response_type="json_object",
-            user=user,
-            query_images=query_images,
-            tracer=tracer,
+    try:
+        with timer("Chat actor: Infer information sources to refer", logger):
+            response = await send_message_to_model_wrapper(
+                query=query,
+                context=function_planning_prompt,
+                response_type="json_object",
+                user=user,
+                query_images=query_images,
+                query_files=query_files,
+                tracer=tracer,
+            )
+    except Exception as e:
+        logger.error(f"Failed to infer information sources to refer: {e}", exc_info=True)
+        yield InformationCollectionIteration(
+            tool=None,
+            query=None,
+            warning="Failed to infer information sources to refer. Skipping iteration. Try again.",
         )
+        return
 
     try:
         response = clean_json(response)
@@ -103,8 +110,15 @@ async def apick_next_tool(
         selected_tool = response.get("tool", None)
         generated_query = response.get("query", None)
         scratchpad = response.get("scratchpad", None)
+        warning = None
         logger.info(f"Response for determining relevant tools: {response}")
-        if send_status_func:
+
+        # Detect selection of previously used query, tool combination.
+        previous_tool_query_combinations = {(i.tool, i.query) for i in previous_iterations}
+        if (selected_tool, generated_query) in previous_tool_query_combinations:
+            warning = f"Repeated tool, query combination detected. Skipping iteration. Try something different."
+        # Only send client status updates if we'll execute this iteration
+        elif send_status_func:
             determined_tool_message = "**Determined Tool**: "
             determined_tool_message += f"{selected_tool}({generated_query})." if selected_tool else "respond."
             determined_tool_message += f"\nReason: {scratchpad}" if scratchpad else ""
@@ -114,13 +128,14 @@ async def apick_next_tool(
         yield InformationCollectionIteration(
             tool=selected_tool,
             query=generated_query,
+            warning=warning,
         )
-
     except Exception as e:
         logger.error(f"Invalid response for determining relevant tools: {response}. {e}", exc_info=True)
         yield InformationCollectionIteration(
             tool=None,
             query=None,
+            warning=f"Invalid response for determining relevant tools: {response}. Skipping iteration. Fix error: {e}",
         )
 
 
@@ -137,6 +152,7 @@ async def execute_information_collection(
     location: LocationData = None,
     file_filters: List[str] = [],
     tracer: dict = {},
+    query_files: str = None,
 ):
     current_iteration = 0
     MAX_ITERATIONS = 5
@@ -147,7 +163,6 @@ async def execute_information_collection(
         document_results: List[Dict[str, str]] = []
         summarize_files: str = ""
         this_iteration = InformationCollectionIteration(tool=None, query=query)
-        previous_iterations_history = construct_iteration_history(previous_iterations, prompts.previous_iteration)
 
         async for result in apick_next_tool(
             query,
@@ -157,19 +172,27 @@ async def execute_information_collection(
             location,
             user_name,
             agent,
-            previous_iterations_history,
+            previous_iterations,
             MAX_ITERATIONS,
             send_status_func,
             tracer=tracer,
+            query_files=query_files,
         ):
             if isinstance(result, dict) and ChatEvent.STATUS in result:
                 yield result[ChatEvent.STATUS]
             elif isinstance(result, InformationCollectionIteration):
                 this_iteration = result
 
-        if this_iteration.tool == ConversationCommand.Notes:
+        # Skip running iteration if warning present in iteration
+        if this_iteration.warning:
+            logger.warning(f"Research mode: {this_iteration.warning}.")
+
+        elif this_iteration.tool == ConversationCommand.Notes:
             this_iteration.context = []
             document_results = []
+            previous_inferred_queries = {
+                c["query"] for iteration in previous_iterations if iteration.context for c in iteration.context
+            }
             async for result in extract_references_and_questions(
                 request,
                 construct_tool_chat_history(previous_iterations, ConversationCommand.Notes),
@@ -181,8 +204,10 @@ async def execute_information_collection(
                 location,
                 send_status_func,
                 query_images,
+                previous_inferred_queries=previous_inferred_queries,
                 agent=agent,
                 tracer=tracer,
+                query_files=query_files,
             ):
                 if isinstance(result, dict) and ChatEvent.STATUS in result:
                     yield result[ChatEvent.STATUS]
@@ -204,6 +229,12 @@ async def execute_information_collection(
                     logger.error(f"Error extracting document references: {e}", exc_info=True)
 
         elif this_iteration.tool == ConversationCommand.Online:
+            previous_subqueries = {
+                subquery
+                for iteration in previous_iterations
+                if iteration.onlineContext
+                for subquery in iteration.onlineContext.keys()
+            }
             async for result in search_online(
                 this_iteration.query,
                 construct_tool_chat_history(previous_iterations, ConversationCommand.Online),
@@ -213,11 +244,16 @@ async def execute_information_collection(
                 [],
                 max_webpages_to_read=0,
                 query_images=query_images,
+                previous_subqueries=previous_subqueries,
                 agent=agent,
                 tracer=tracer,
             ):
                 if isinstance(result, dict) and ChatEvent.STATUS in result:
                     yield result[ChatEvent.STATUS]
+                elif is_none_or_empty(result):
+                    this_iteration.warning = (
+                        "Detected previously run online search queries. Skipping iteration. Try something different."
+                    )
                 else:
                     online_results: Dict[str, Dict] = result  # type: ignore
                     this_iteration.onlineContext = online_results
@@ -233,6 +269,7 @@ async def execute_information_collection(
                     query_images=query_images,
                     agent=agent,
                     tracer=tracer,
+                    query_files=query_files,
                 ):
                     if isinstance(result, dict) and ChatEvent.STATUS in result:
                         yield result[ChatEvent.STATUS]
@@ -263,6 +300,7 @@ async def execute_information_collection(
                     send_status_func,
                     query_images=query_images,
                     agent=agent,
+                    query_files=query_files,
                     tracer=tracer,
                 ):
                     if isinstance(result, dict) and ChatEvent.STATUS in result:
@@ -288,6 +326,7 @@ async def execute_information_collection(
                     query_images=query_images,
                     agent=agent,
                     send_status_func=send_status_func,
+                    query_files=query_files,
                 ):
                     if isinstance(result, dict) and ChatEvent.STATUS in result:
                         yield result[ChatEvent.STATUS]
@@ -302,16 +341,19 @@ async def execute_information_collection(
 
         current_iteration += 1
 
-        if document_results or online_results or code_results or summarize_files:
-            results_data = f"**Results**:\n"
+        if document_results or online_results or code_results or summarize_files or this_iteration.warning:
+            results_data = f"\n<iteration>{current_iteration}\n<tool>{this_iteration.tool}</tool>\n<query>{this_iteration.query}</query>\n<results>"
             if document_results:
-                results_data += f"**Document References**:\n{yaml.dump(document_results, allow_unicode=True, sort_keys=False, default_flow_style=False)}\n"
+                results_data += f"\n<document_references>\n{yaml.dump(document_results, allow_unicode=True, sort_keys=False, default_flow_style=False)}\n</document_references>"
             if online_results:
-                results_data += f"**Online Results**:\n{yaml.dump(online_results, allow_unicode=True, sort_keys=False, default_flow_style=False)}\n"
+                results_data += f"\n<online_results>\n{yaml.dump(online_results, allow_unicode=True, sort_keys=False, default_flow_style=False)}\n</online_results>"
             if code_results:
-                results_data += f"**Code Results**:\n{yaml.dump(code_results, allow_unicode=True, sort_keys=False, default_flow_style=False)}\n"
+                results_data += f"\n<code_results>\n{yaml.dump(code_results, allow_unicode=True, sort_keys=False, default_flow_style=False)}\n</code_results>"
             if summarize_files:
-                results_data += f"**Summarized Files**:\n{yaml.dump(summarize_files, allow_unicode=True, sort_keys=False, default_flow_style=False)}\n"
+                results_data += f"\n<summarized_files>\n{yaml.dump(summarize_files, allow_unicode=True, sort_keys=False, default_flow_style=False)}\n</summarized_files>"
+            if this_iteration.warning:
+                results_data += f"\n<warning>\n{this_iteration.warning}\n</warning>"
+            results_data += "\n</results>\n</iteration>"
 
             # intermediate_result = await extract_relevant_info(this_iteration.query, results_data, agent)
             this_iteration.summarizedResult = results_data

khoj/routers/web_client.py

@@ -51,16 +51,6 @@ def chat_page(request: Request):
     )
 
 
-@web_client.get("/factchecker", response_class=FileResponse)
-def fact_checker_page(request: Request):
-    return templates.TemplateResponse(
-        "factchecker/index.html",
-        context={
-            "request": request,
-        },
-    )
-
-
 @web_client.get("/login", response_class=FileResponse)
 def login_page(request: Request):
     next_url = get_next_url(request)

khoj/search_type/text_search.py

@@ -8,11 +8,7 @@ import torch
 from asgiref.sync import sync_to_async
 from sentence_transformers import util
 
-from khoj.database.adapters import (
-    EntryAdapters,
-    get_default_search_model,
-    get_user_default_search_model,
-)
+from khoj.database.adapters import EntryAdapters, get_default_search_model
 from khoj.database.models import Agent
 from khoj.database.models import Entry as DbEntry
 from khoj.database.models import KhojUser
@@ -114,7 +110,7 @@ async def query(
     file_type = search_type_to_embeddings_type[type.value]
 
     query = raw_query
-    search_model = await sync_to_async(get_user_default_search_model)(user)
+    search_model = await sync_to_async(get_default_search_model)()
     if not max_distance:
         if search_model.bi_encoder_confidence_threshold:
             max_distance = search_model.bi_encoder_confidence_threshold
@@ -212,7 +208,7 @@ def setup(
     text_to_entries: Type[TextToEntries],
     files: dict[str, str],
     regenerate: bool,
-    user: KhojUser = None,
+    user: KhojUser,
     config=None,
 ) -> Tuple[int, int]:
     if config:

khoj/utils/cli.py

@@ -16,7 +16,7 @@ from khoj.migrations.migrate_processor_config_openai import (
 )
 from khoj.migrations.migrate_server_pg import migrate_server_pg
 from khoj.migrations.migrate_version import migrate_config_to_version
-from khoj.utils.helpers import in_debug_mode, resolve_absolute_path
+from khoj.utils.helpers import in_debug_mode, is_env_var_true, resolve_absolute_path
 from khoj.utils.yaml import parse_config_from_file
 
 
@@ -79,7 +79,7 @@ def cli(args=None):
     else:
         args = run_migrations(args)
         args.config = parse_config_from_file(args.config_file)
-        if in_debug_mode():
+        if is_env_var_true("KHOJ_TELEMETRY_DISABLE") or in_debug_mode():
             args.config.app.should_log_telemetry = False
 
     return args

khoj/utils/fs_syncer.py

@@ -8,6 +8,7 @@ from bs4 import BeautifulSoup
 from magika import Magika
 
 from khoj.database.models import (
+    KhojUser,
     LocalMarkdownConfig,
     LocalOrgConfig,
     LocalPdfConfig,
@@ -21,7 +22,7 @@ logger = logging.getLogger(__name__)
 magika = Magika()
 
 
-def collect_files(search_type: Optional[SearchType] = SearchType.All, user=None) -> dict:
+def collect_files(user: KhojUser, search_type: Optional[SearchType] = SearchType.All) -> dict:
     files: dict[str, dict] = {"docx": {}, "image": {}}
 
     if search_type == SearchType.All or search_type == SearchType.Org:

khoj/utils/helpers.py

@@ -254,8 +254,10 @@ def get_server_id():
     return server_id
 
 
-def telemetry_disabled(app_config: AppConfig):
-    return not app_config or not app_config.should_log_telemetry
+def telemetry_disabled(app_config: AppConfig, telemetry_disable_env) -> bool:
+    return (
+        not app_config.should_log_telemetry if app_config and app_config.should_log_telemetry else telemetry_disable_env
+    )
 
 
 def log_telemetry(
@@ -263,11 +265,12 @@ def log_telemetry(
     api: str = None,
     client: Optional[str] = None,
     app_config: Optional[AppConfig] = None,
+    disable_telemetry_env: bool = False,
     properties: dict = None,
 ):
     """Log basic app usage telemetry like client, os, api called"""
     # Do not log usage telemetry, if telemetry is disabled via app config
-    if telemetry_disabled(app_config):
+    if telemetry_disabled(app_config, disable_telemetry_env):
         return []
 
     if properties.get("server_id") is None:

khoj/utils/rawconfig.py

@@ -138,6 +138,38 @@ class SearchResponse(ConfigBase):
     corpus_id: str
 
 
+class FileData(BaseModel):
+    name: str
+    content: bytes
+    file_type: str
+    encoding: str | None = None
+
+
+class FileAttachment(BaseModel):
+    name: str
+    content: str
+    file_type: str
+    size: int
+
+
+class ChatRequestBody(BaseModel):
+    q: str
+    n: Optional[int] = 7
+    d: Optional[float] = None
+    stream: Optional[bool] = False
+    title: Optional[str] = None
+    conversation_id: Optional[str] = None
+    turn_id: Optional[str] = None
+    city: Optional[str] = None
+    region: Optional[str] = None
+    country: Optional[str] = None
+    country_code: Optional[str] = None
+    timezone: Optional[str] = None
+    images: Optional[list[str]] = None
+    files: Optional[list[FileAttachment]] = []
+    create_new: Optional[bool] = False
+
+
 class Entry:
     raw: str
     compiled: str

khoj/utils/state.py

@@ -12,7 +12,7 @@ from khoj.database.models import ProcessLock
 from khoj.processor.embeddings import CrossEncoderModel, EmbeddingsModel
 from khoj.utils import config as utils_config
 from khoj.utils.config import OfflineChatProcessorModel, SearchModels
-from khoj.utils.helpers import LRU, get_device
+from khoj.utils.helpers import LRU, get_device, is_env_var_true
 from khoj.utils.rawconfig import FullConfig
 
 # Application Global State
@@ -34,6 +34,7 @@ SearchType = utils_config.SearchType
 scheduler: BackgroundScheduler = None
 schedule_leader_process_lock: ProcessLock = None
 telemetry: List[Dict[str, str]] = []
+telemetry_disabled: bool = is_env_var_true("KHOJ_TELEMETRY_DISABLE")
 khoj_version: str = None
 device = get_device()
 chat_on_gpu: bool = True

{khoj-1.28.3.dist-info → khoj-1.28.4.dev92.dist-info}/METADATA

@@ -1,13 +1,11 @@
 Metadata-Version: 2.3
 Name: khoj
-Version: 1.28.3
+Version: 1.28.4.dev92
 Summary: Your Second Brain
 Project-URL: Homepage, https://khoj.dev
 Project-URL: Documentation, https://docs.khoj.dev
 Project-URL: Code, https://github.com/khoj-ai/khoj
 Author: Debanjum Singh Solanky, Saba Imran
-License-Expression: AGPL-3.0-or-later
-License-File: LICENSE
 Keywords: AI,NLP,images,markdown,org-mode,pdf,productivity,search,semantic-search
 Classifier: Development Status :: 5 - Production/Stable
 Classifier: Intended Audience :: Information Technology
@@ -76,12 +74,14 @@ Requires-Dist: websockets==12.0
 Provides-Extra: dev
 Requires-Dist: black>=23.1.0; extra == 'dev'
 Requires-Dist: boto3>=1.34.57; extra == 'dev'
+Requires-Dist: datasets; extra == 'dev'
 Requires-Dist: factory-boy>=3.2.1; extra == 'dev'
 Requires-Dist: freezegun>=1.2.0; extra == 'dev'
 Requires-Dist: gitpython~=3.1.43; extra == 'dev'
 Requires-Dist: google-auth==2.23.3; extra == 'dev'
 Requires-Dist: gunicorn==22.0.0; extra == 'dev'
 Requires-Dist: mypy>=1.0.1; extra == 'dev'
+Requires-Dist: pandas; extra == 'dev'
 Requires-Dist: pre-commit>=3.0.4; extra == 'dev'
 Requires-Dist: pytest-asyncio==0.21.1; extra == 'dev'
 Requires-Dist: pytest-django==4.5.2; extra == 'dev'