khoj 1.40.1.dev1__py3-none-any.whl → 1.40.1.dev13__py3-none-any.whl

khoj/processor/conversation/openai/utils.py

@@ -64,9 +64,13 @@ def completion_with_backoff(
     formatted_messages = [{"role": message.role, "content": message.content} for message in messages]
 
     # Tune reasoning models arguments
-    if model_name.startswith("o1") or model_name.startswith("o3"):
+    if is_openai_reasoning_model(model_name, api_base_url):
         temperature = 1
-        model_kwargs["reasoning_effort"] = "medium"
+        reasoning_effort = "medium" if deepthought else "low"
+        model_kwargs["reasoning_effort"] = reasoning_effort
+    elif is_twitter_reasoning_model(model_name, api_base_url):
+        reasoning_effort = "high" if deepthought else "low"
+        model_kwargs["reasoning_effort"] = reasoning_effort
 
     model_kwargs["stream_options"] = {"include_usage": True}
     if os.getenv("KHOJ_LLM_SEED"):
@@ -162,12 +166,13 @@ def llm_thread(
 
         formatted_messages = [{"role": message.role, "content": message.content} for message in messages]
 
-        # Tune reasoning models arguments
-        if model_name.startswith("o1") or model_name.startswith("o3"):
+        # Configure thinking for openai reasoning models
+        if is_openai_reasoning_model(model_name, api_base_url):
             temperature = 1
-            model_kwargs["reasoning_effort"] = "medium"
+            reasoning_effort = "medium" if deepthought else "low"
+            model_kwargs["reasoning_effort"] = reasoning_effort
+            model_kwargs.pop("stop", None)  # Remove unsupported stop param for reasoning models
 
-        if model_name.startswith("o3"):
             # Get the first system message and add the string `Formatting re-enabled` to it.
             # See https://platform.openai.com/docs/guides/reasoning-best-practices
             if len(formatted_messages) > 0:
@@ -179,6 +184,9 @@ def llm_thread(
                     formatted_messages[first_system_message_index][
                         "content"
                     ] = f"{first_system_message} Formatting re-enabled"
+        elif is_twitter_reasoning_model(model_name, api_base_url):
+            reasoning_effort = "high" if deepthought else "low"
+            model_kwargs["reasoning_effort"] = reasoning_effort
         elif model_name.startswith("deepseek-reasoner"):
             # Two successive messages cannot be from the same role. Should merge any back-to-back messages from the same role.
             # The first message should always be a user message (except system message).
@@ -257,3 +265,21 @@ def get_openai_api_json_support(model_name: str, api_base_url: str = None) -> Js
         if host == "api.deepinfra.com":
             return JsonSupport.OBJECT
     return JsonSupport.SCHEMA
+
+
+def is_openai_reasoning_model(model_name: str, api_base_url: str = None) -> bool:
+    """
+    Check if the model is an OpenAI reasoning model
+    """
+    return model_name.startswith("o") and (api_base_url is None or api_base_url.startswith("https://api.openai.com/v1"))
+
+
+def is_twitter_reasoning_model(model_name: str, api_base_url: str = None) -> bool:
+    """
+    Check if the model is a Twitter reasoning model
+    """
+    return (
+        model_name.startswith("grok-3-mini")
+        and api_base_url is not None
+        and api_base_url.startswith("https://api.x.ai/v1")
+    )

khoj/processor/tools/run_code.py

@@ -244,6 +244,7 @@ async def execute_e2b(code: str, input_files: list[dict]) -> dict[str, Any]:
 
         # Collect output files
         output_files = []
+        image_file_ext = {".png", ".jpeg", ".jpg", ".svg"}
 
         # Identify new files created during execution
         new_files = set(E2bFile(f.name, f.path) for f in await sandbox.files.list("~")) - original_files
@@ -254,7 +255,7 @@ async def execute_e2b(code: str, input_files: list[dict]) -> dict[str, Any]:
             if isinstance(content, bytes):
                 # Binary files like PNG - encode as base64
                 b64_data = base64.b64encode(content).decode("utf-8")
-            elif Path(f.name).suffix in [".png", ".jpeg", ".jpg", ".svg"]:
+            elif Path(f.name).suffix in image_file_ext:
                 # Ignore image files as they are extracted from execution results below for inline display
                 continue
             else:
@@ -263,8 +264,12 @@ async def execute_e2b(code: str, input_files: list[dict]) -> dict[str, Any]:
             output_files.append({"filename": f.name, "b64_data": b64_data})
 
         # Collect output files from execution results
+        # Repect ordering of output result types to disregard text output associated with images
+        output_result_types = ["png", "jpeg", "svg", "text", "markdown", "json"]
         for idx, result in enumerate(execution.results):
-            for result_type in {"png", "jpeg", "svg", "text", "markdown", "json"}:
+            if getattr(result, "chart", None):
+                continue
+            for result_type in output_result_types:
                 if b64_data := getattr(result, result_type, None):
                     output_files.append({"filename": f"{idx}.{result_type}", "b64_data": b64_data})
                     break

khoj/routers/api_agents.py

@@ -12,7 +12,7 @@ from pydantic import BaseModel
 from starlette.authentication import has_required_scope, requires
 
 from khoj.database.adapters import AgentAdapters, ConversationAdapters, EntryAdapters
-from khoj.database.models import Agent, Conversation, KhojUser
+from khoj.database.models import Agent, Conversation, KhojUser, PriceTier
 from khoj.routers.helpers import CommonQueryParams, acheck_if_safe_prompt
 from khoj.utils.helpers import (
     ConversationCommand,
@@ -125,8 +125,20 @@ async def get_agent_by_conversation(
     else:
         agent = await AgentAdapters.aget_default_agent()
 
+    if agent is None:
+        return Response(
+            content=json.dumps({"error": f"Agent for conversation id {conversation_id} not found for user {user}."}),
+            media_type="application/json",
+            status_code=404,
+        )
+
+    chat_model = await AgentAdapters.aget_agent_chat_model(agent, user)
+    if is_subscribed or chat_model.price_tier == PriceTier.FREE:
+        agent_chat_model = chat_model.name
+    else:
+        agent_chat_model = None
+
     has_files = agent.fileobject_set.exists()
-    agent.chat_model = await AgentAdapters.aget_agent_chat_model(agent, user)
 
     agents_packet = {
         "slug": agent.slug,
@@ -137,7 +149,7 @@ async def get_agent_by_conversation(
         "color": agent.style_color,
         "icon": agent.style_icon,
         "privacy_level": agent.privacy_level,
-        "chat_model": agent.chat_model.name if is_subscribed else None,
+        "chat_model": agent_chat_model,
         "has_files": has_files,
         "input_tools": agent.input_tools,
         "output_modes": agent.output_modes,
@@ -249,7 +261,11 @@ async def update_hidden_agent(
     user: KhojUser = request.user.object
 
     subscribed = has_required_scope(request, ["premium"])
-    chat_model = body.chat_model if subscribed else None
+    chat_model = await ConversationAdapters.aget_chat_model_by_name(body.chat_model)
+    if subscribed or chat_model.price_tier == PriceTier.FREE:
+        agent_chat_model = body.chat_model
+    else:
+        agent_chat_model = None
 
     selected_agent = await AgentAdapters.aget_agent_by_slug(body.slug, user)
 
@@ -264,7 +280,7 @@ async def update_hidden_agent(
         user=user,
         slug=body.slug,
         persona=body.persona,
-        chat_model=chat_model,
+        chat_model=agent_chat_model,
         input_tools=body.input_tools,
         output_modes=body.output_modes,
         existing_agent=selected_agent,
@@ -295,7 +311,11 @@ async def create_hidden_agent(
     user: KhojUser = request.user.object
 
     subscribed = has_required_scope(request, ["premium"])
-    chat_model = body.chat_model if subscribed else None
+    chat_model = await ConversationAdapters.aget_chat_model_by_name(body.chat_model)
+    if subscribed or chat_model.price_tier == PriceTier.FREE:
+        agent_chat_model = body.chat_model
+    else:
+        agent_chat_model = None
 
     conversation = await ConversationAdapters.aget_conversation_by_user(user=user, conversation_id=conversation_id)
     if not conversation:
@@ -320,7 +340,7 @@ async def create_hidden_agent(
         user=user,
         slug=body.slug,
         persona=body.persona,
-        chat_model=chat_model,
+        chat_model=agent_chat_model,
         input_tools=body.input_tools,
         output_modes=body.output_modes,
         existing_agent=None,
@@ -364,7 +384,11 @@ async def create_agent(
         )
 
     subscribed = has_required_scope(request, ["premium"])
-    chat_model = body.chat_model if subscribed else None
+    chat_model = await ConversationAdapters.aget_chat_model_by_name(body.chat_model)
+    if subscribed or chat_model.price_tier == PriceTier.FREE:
+        agent_chat_model = body.chat_model
+    else:
+        agent_chat_model = None
 
     agent = await AgentAdapters.aupdate_agent(
         user,
@@ -373,7 +397,7 @@ async def create_agent(
         body.privacy_level,
         body.icon,
         body.color,
-        chat_model,
+        agent_chat_model,
         body.files,
         body.input_tools,
         body.output_modes,
@@ -431,7 +455,11 @@ async def update_agent(
         )
 
     subscribed = has_required_scope(request, ["premium"])
-    chat_model = body.chat_model if subscribed else None
+    chat_model = await ConversationAdapters.aget_chat_model_by_name(body.chat_model)
+    if subscribed or chat_model.price_tier == PriceTier.FREE:
+        agent_chat_model = body.chat_model
+    else:
+        agent_chat_model = None
 
     agent = await AgentAdapters.aupdate_agent(
         user,
@@ -440,7 +468,7 @@ async def update_agent(
         body.privacy_level,
         body.icon,
         body.color,
-        chat_model,
+        agent_chat_model,
         body.files,
         body.input_tools,
         body.output_modes,

khoj/routers/api_model.py

@@ -2,13 +2,18 @@ import json
 import logging
 from typing import Dict, Optional, Union
 
-from fastapi import APIRouter, HTTPException, Request
+from fastapi import APIRouter, Request
 from fastapi.requests import Request
 from fastapi.responses import Response
 from starlette.authentication import has_required_scope, requires
 
-from khoj.database import adapters
-from khoj.database.adapters import ConversationAdapters, EntryAdapters
+from khoj.database.adapters import ConversationAdapters
+from khoj.database.models import (
+    ChatModel,
+    PriceTier,
+    TextToImageModelConfig,
+    VoiceModelOption,
+)
 from khoj.routers.helpers import update_telemetry_state
 
 api_model = APIRouter()
@@ -53,13 +58,24 @@ def get_user_chat_model(
 
 
 @api_model.post("/chat", status_code=200)
-@requires(["authenticated", "premium"])
+@requires(["authenticated"])
 async def update_chat_model(
     request: Request,
     id: str,
     client: Optional[str] = None,
 ):
     user = request.user.object
+    subscribed = has_required_scope(request, ["premium"])
+
+    # Validate if model can be switched
+    chat_model = await ChatModel.objects.filter(id=int(id)).afirst()
+    if chat_model is None:
+        return Response(status_code=404, content=json.dumps({"status": "error", "message": "Chat model not found"}))
+    if not subscribed and chat_model.price_tier != PriceTier.FREE:
+        raise Response(
+            status_code=403,
+            content=json.dumps({"status": "error", "message": "Subscribe to switch to this chat model"}),
+        )
 
     new_config = await ConversationAdapters.aset_user_conversation_processor(user, int(id))
 
@@ -78,13 +94,24 @@ async def update_chat_model(
 
 
 @api_model.post("/voice", status_code=200)
-@requires(["authenticated", "premium"])
+@requires(["authenticated"])
 async def update_voice_model(
     request: Request,
     id: str,
     client: Optional[str] = None,
 ):
     user = request.user.object
+    subscribed = has_required_scope(request, ["premium"])
+
+    # Validate if model can be switched
+    voice_model = await VoiceModelOption.objects.filter(id=int(id)).afirst()
+    if voice_model is None:
+        return Response(status_code=404, content=json.dumps({"status": "error", "message": "Voice model not found"}))
+    if not subscribed and voice_model.price_tier != PriceTier.FREE:
+        raise Response(
+            status_code=403,
+            content=json.dumps({"status": "error", "message": "Subscribe to switch to this voice model"}),
+        )
 
     new_config = await ConversationAdapters.aset_user_voice_model(user, id)
 
@@ -111,8 +138,15 @@ async def update_paint_model(
     user = request.user.object
     subscribed = has_required_scope(request, ["premium"])
 
-    if not subscribed:
-        raise HTTPException(status_code=403, detail="User is not subscribed to premium")
+    # Validate if model can be switched
+    image_model = await TextToImageModelConfig.objects.filter(id=int(id)).afirst()
+    if image_model is None:
+        return Response(status_code=404, content=json.dumps({"status": "error", "message": "Image model not found"}))
+    if not subscribed and image_model.price_tier != PriceTier.FREE:
+        raise Response(
+            status_code=403,
+            content=json.dumps({"status": "error", "message": "Subscribe to switch to this image model"}),
+        )
 
     new_config = await ConversationAdapters.aset_user_text_to_image_model(user, int(id))
 

khoj/routers/helpers.py

@@ -1290,6 +1290,7 @@ async def send_message_to_model_wrapper(
             model=chat_model_name,
             response_type=response_type,
             response_schema=response_schema,
+            deepthought=deepthought,
             api_base_url=api_base_url,
             tracer=tracer,
         )
@@ -1593,6 +1594,7 @@ def generate_chat_response(
                 generated_files=raw_generated_files,
                 generated_asset_results=generated_asset_results,
                 program_execution_context=program_execution_context,
+                deepthought=deepthought,
                 tracer=tracer,
             )
 
@@ -2362,6 +2364,7 @@ def get_user_config(user: KhojUser, request: Request, is_detailed: bool = False)
                 "id": chat_model.id,
                 "strengths": chat_model.strengths,
                 "description": chat_model.description,
+                "tier": chat_model.price_tier,
             }
         )
 
@@ -2369,12 +2372,24 @@ def get_user_config(user: KhojUser, request: Request, is_detailed: bool = False)
     paint_model_options = ConversationAdapters.get_text_to_image_model_options().all()
     all_paint_model_options = list()
     for paint_model in paint_model_options:
-        all_paint_model_options.append({"name": paint_model.model_name, "id": paint_model.id})
+        all_paint_model_options.append(
+            {
+                "name": paint_model.model_name,
+                "id": paint_model.id,
+                "tier": paint_model.price_tier,
+            }
+        )
 
     voice_models = ConversationAdapters.get_voice_model_options()
     voice_model_options = list()
     for voice_model in voice_models:
-        voice_model_options.append({"name": voice_model.name, "id": voice_model.model_id})
+        voice_model_options.append(
+            {
+                "name": voice_model.name,
+                "id": voice_model.model_id,
+                "tier": voice_model.price_tier,
+            }
+        )
 
     if len(voice_model_options) == 0:
         eleven_labs_enabled = False

khoj/utils/constants.py

@@ -39,14 +39,18 @@ model_to_cost: Dict[str, Dict[str, float]] = {
     "gpt-4o": {"input": 2.50, "output": 10.00},
     "gpt-4o-mini": {"input": 0.15, "output": 0.60},
     "o1": {"input": 15.0, "output": 60.00},
+    "o3": {"input": 10.0, "output": 40.00},
     "o1-mini": {"input": 3.0, "output": 12.0},
     "o3-mini": {"input": 1.10, "output": 4.40},
+    "o4-mini": {"input": 1.10, "output": 4.40},
     # Gemini Pricing: https://ai.google.dev/pricing
     "gemini-1.5-flash": {"input": 0.075, "output": 0.30},
     "gemini-1.5-flash-002": {"input": 0.075, "output": 0.30},
     "gemini-1.5-pro": {"input": 1.25, "output": 5.00},
     "gemini-1.5-pro-002": {"input": 1.25, "output": 5.00},
     "gemini-2.0-flash": {"input": 0.10, "output": 0.40},
+    "gemini-2.5-flash-preview-04-17": {"input": 0.15, "output": 0.60, "thought": 3.50},
+    "gemini-2.5-pro-preview-03-25": {"input": 1.25, "output": 10.0},
     # Anthropic Pricing: https://www.anthropic.com/pricing#anthropic-api
     "claude-3-5-haiku-20241022": {"input": 1.0, "output": 5.0, "cache_read": 0.08, "cache_write": 1.0},
     "claude-3-5-haiku@20241022": {"input": 1.0, "output": 5.0, "cache_read": 0.08, "cache_write": 1.0},
@@ -55,4 +59,9 @@ model_to_cost: Dict[str, Dict[str, float]] = {
     "claude-3-7-sonnet-20250219": {"input": 3.0, "output": 15.0, "cache_read": 0.3, "cache_write": 3.75},
     "claude-3-7-sonnet@20250219": {"input": 3.0, "output": 15.0, "cache_read": 0.3, "cache_write": 3.75},
     "claude-3-7-sonnet-latest": {"input": 3.0, "output": 15.0, "cache_read": 0.3, "cache_write": 3.75},
+    # Grok pricing: https://docs.x.ai/docs/models
+    "grok-3": {"input": 3.0, "output": 15.0},
+    "grok-3-latest": {"input": 3.0, "output": 15.0},
+    "grok-3-mini": {"input": 0.30, "output": 0.50},
+    "grok-3-mini-latest": {"input": 0.30, "output": 0.50},
 }

khoj/utils/helpers.py

@@ -601,6 +601,7 @@ def get_cost_of_chat_message(
     model_name: str,
     input_tokens: int = 0,
     output_tokens: int = 0,
+    thought_tokens: int = 0,
     cache_read_tokens: int = 0,
     cache_write_tokens: int = 0,
     prev_cost: float = 0.0,
@@ -612,10 +613,11 @@ def get_cost_of_chat_message(
     # Calculate cost of input and output tokens. Costs are per million tokens
     input_cost = constants.model_to_cost.get(model_name, {}).get("input", 0) * (input_tokens / 1e6)
     output_cost = constants.model_to_cost.get(model_name, {}).get("output", 0) * (output_tokens / 1e6)
+    thought_cost = constants.model_to_cost.get(model_name, {}).get("thought", 0) * (thought_tokens / 1e6)
     cache_read_cost = constants.model_to_cost.get(model_name, {}).get("cache_read", 0) * (cache_read_tokens / 1e6)
     cache_write_cost = constants.model_to_cost.get(model_name, {}).get("cache_write", 0) * (cache_write_tokens / 1e6)
 
-    return input_cost + output_cost + cache_read_cost + cache_write_cost + prev_cost
+    return input_cost + output_cost + thought_cost + cache_read_cost + cache_write_cost + prev_cost
 
 
 def get_chat_usage_metrics(
@@ -624,6 +626,7 @@ def get_chat_usage_metrics(
     output_tokens: int = 0,
     cache_read_tokens: int = 0,
     cache_write_tokens: int = 0,
+    thought_tokens: int = 0,
     usage: dict = {},
     cost: float = None,
 ):
@@ -633,6 +636,7 @@ def get_chat_usage_metrics(
     prev_usage = usage or {
         "input_tokens": 0,
         "output_tokens": 0,
+        "thought_tokens": 0,
         "cache_read_tokens": 0,
         "cache_write_tokens": 0,
         "cost": 0.0,
@@ -640,11 +644,18 @@ def get_chat_usage_metrics(
     return {
         "input_tokens": prev_usage["input_tokens"] + input_tokens,
         "output_tokens": prev_usage["output_tokens"] + output_tokens,
+        "thought_tokens": prev_usage.get("thought_tokens", 0) + thought_tokens,
         "cache_read_tokens": prev_usage.get("cache_read_tokens", 0) + cache_read_tokens,
         "cache_write_tokens": prev_usage.get("cache_write_tokens", 0) + cache_write_tokens,
         "cost": cost
         or get_cost_of_chat_message(
-            model_name, input_tokens, output_tokens, cache_read_tokens, cache_write_tokens, prev_cost=prev_usage["cost"]
+            model_name,
+            input_tokens,
+            output_tokens,
+            thought_tokens,
+            cache_read_tokens,
+            cache_write_tokens,
+            prev_cost=prev_usage["cost"],
         ),
     }
 

{khoj-1.40.1.dev1.dist-info → khoj-1.40.1.dev13.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: khoj
-Version: 1.40.1.dev1
+Version: 1.40.1.dev13
 Summary: Your Second Brain
 Project-URL: Homepage, https://khoj.dev
 Project-URL: Documentation, https://docs.khoj.dev
@@ -40,7 +40,7 @@ Requires-Dist: einops==0.8.0
 Requires-Dist: email-validator==2.2.0
 Requires-Dist: fastapi>=0.110.0
 Requires-Dist: google-auth~=2.23.3
-Requires-Dist: google-genai==1.5.0
+Requires-Dist: google-genai==1.11.0
 Requires-Dist: httpx==0.28.1
 Requires-Dist: huggingface-hub>=0.22.2
 Requires-Dist: itsdangerous==2.1.2
@@ -150,7 +150,7 @@ Description-Content-Type: text/markdown
 
 [Khoj](https://khoj.dev) is a personal AI app to extend your capabilities. It smoothly scales up from an on-device personal AI to a cloud-scale enterprise AI.
 
-- Chat with any local or online LLM (e.g llama3, qwen, gemma, mistral, gpt, claude, gemini).
+- Chat with any local or online LLM (e.g llama3, qwen, gemma, mistral, gpt, claude, gemini, deepseek).
 - Get answers from the internet and your docs (including image, pdf, markdown, org-mode, word, notion files).
 - Access it from your Browser, Obsidian, Emacs, Desktop, Phone or Whatsapp.
 - Create agents with custom knowledge, persona, chat model and tools to take on any role.
@@ -179,6 +179,19 @@ To get started with self-hosting Khoj, [read the docs](https://docs.khoj.dev/get
 
 Khoj is available as a cloud service, on-premises, or as a hybrid solution. To learn more about Khoj Enterprise, [visit our website](https://khoj.dev/teams).
 
+## Frequently Asked Questions (FAQ)
+
+Q: Can I use Khoj without self-hosting?
+Yes! You can use Khoj right away at [https://app.khoj.dev](https://app.khoj.dev) — no setup required.
+
+Q: What kinds of documents can Khoj read?
+Khoj supports a wide variety: PDFs, Markdown, Notion, Word docs, org-mode files, and more.
+
+Q: How can I make my own agent?
+Check out [this blog post](https://blog.khoj.dev/posts/create-agents-on-khoj/) for a step-by-step guide to custom agents.
+For more questions, head over to our [Discord](https://discord.gg/BDgyabRM6e)!
+
+
 ## Contributors
 Cheers to our awesome contributors! 🎉