klbr/khoj
1
0
Fork 0
mirror of https://github.com/khoaliber/khoj.git synced 2026-03-05 05:39:11 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'master' into features/advanced-reasoning

Browse Source
This commit is contained in:
Debanjum
2024-10-28 04:07:36 -07:00
parent a5d0ca6e1c aad7528d1b
commit 50ffd7f199
12 changed files with 116 additions and 96 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
src/interface/web/app/components/chatInputArea/chatInputArea.tsx
View File

@@ -394,6 +394,11 @@ export const ChatInputArea = forwardRef<HTMLTextAreaElement, ChatInputProps>((pr
<PopoverContent
onOpenAutoFocus={(e) => e.preventDefault()}
className={`${props.isMobileWidth ? "w-[100vw]" : "w-full"} rounded-md`}
side="top"
align="center"
/* Offset below text area on home page (i.e where conversationId is unset) */
sideOffset={props.conversationId ? 0 : 80}
alignOffset={0}
>
<Command className="max-w-full">
<CommandInput

2
src/interface/web/app/page.tsx
View File

@@ -362,7 +362,7 @@ function ChatBodyData(props: ChatBodyDataProps) {
className={`${selectedAgent === agents[index].slug ? convertColorToBorderClass(agents[index].color) : "border-muted text-muted-foreground"} hover:cursor-pointer`}
>
<CardTitle
className="text-center text-xs font-medium flex justify-center items-center px-1.5 py-1"
className="text-center text-xs font-medium flex justify-center items-center whitespace-nowrap px-1.5 py-1"
onDoubleClick={() =>
openAgentEditCard(agents[index].slug)
}

32
src/khoj/database/management/commands/change_default_model.py
View File

@@ -19,6 +19,8 @@ from khoj.processor.embeddings import EmbeddingsModel
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)
BATCH_SIZE = 1000 # Define an appropriate batch size
class Command(BaseCommand):
help = "Convert all existing Entry objects to use a new default Search model."
@@ -42,22 +44,24 @@ class Command(BaseCommand):
def handle(self, *args, **options):
@transaction.atomic
def regenerate_entries(entry_filter: Q, embeddings_model: EmbeddingsModel, search_model: SearchModelConfig):
entries = Entry.objects.filter(entry_filter).all()
compiled_entries = [entry.compiled for entry in entries]
updated_entries: List[Entry] = []
try:
embeddings = embeddings_model.embed_documents(compiled_entries)
total_entries = Entry.objects.filter(entry_filter).count()
for start in tqdm(range(0, total_entries, BATCH_SIZE)):
end = start + BATCH_SIZE
entries = Entry.objects.filter(entry_filter)[start:end]
compiled_entries = [entry.compiled for entry in entries]
updated_entries: List[Entry] = []
try:
embeddings = embeddings_model.embed_documents(compiled_entries)
except Exception as e:
logger.error(f"Error embedding documents: {e}")
return
except Exception as e:
logger.error(f"Error embedding documents: {e}")
return
for i, entry in enumerate(entries):
entry.embeddings = embeddings[i]
entry.search_model_id = search_model.id
updated_entries.append(entry)
for i, entry in enumerate(tqdm(entries)):
entry.embeddings = embeddings[i]
entry.search_model_id = search_model.id
updated_entries.append(entry)
Entry.objects.bulk_update(updated_entries, ["embeddings", "search_model_id", "file_path"])
Entry.objects.bulk_update(updated_entries, ["embeddings", "search_model_id", "file_path"])
search_model_config_id = options.get("search_model_id")
apply = options.get("apply")

23
src/khoj/processor/conversation/anthropic/anthropic_chat.py
View File

@@ -152,9 +152,7 @@ def converse_anthropic(
"""
# Initialize Variables
current_date = datetime.now()
compiled_references = "\n\n".join({f"# {item}" for item in references})
conversation_primer = prompts.query_prompt.format(query=user_query)
compiled_references = "\n\n".join({f"# File: {item['file']}\n## {item['compiled']}\n" for item in references})
if agent and agent.personality:
system_prompt = prompts.custom_personality.format(
@@ -185,20 +183,19 @@ def converse_anthropic(
completion_func(chat_response=prompts.no_online_results_found.format())
return iter([prompts.no_online_results_found.format()])
if ConversationCommand.Code in conversation_commands and not is_none_or_empty(code_results):
conversation_primer = (
f"{prompts.code_executed_context.format(code_results=str(code_results))}\n{conversation_primer}"
)
if ConversationCommand.Online in conversation_commands or ConversationCommand.Webpage in conversation_commands:
conversation_primer = (
f"{prompts.online_search_conversation.format(online_results=str(online_results))}\n{conversation_primer}"
)
context_message = ""
if not is_none_or_empty(compiled_references):
conversation_primer = f"{prompts.notes_conversation.format(query=user_query, references=compiled_references)}\n\n{conversation_primer}"
context_message = f"{prompts.notes_conversation.format(query=user_query, references=compiled_references)}\n\n"
if ConversationCommand.Online in conversation_commands or ConversationCommand.Webpage in conversation_commands:
context_message += f"{prompts.online_search_conversation.format(online_results=str(online_results))}\n\n"
if ConversationCommand.Code in conversation_commands and not is_none_or_empty(code_results):
context_message += f"{prompts.code_executed_context.format(code_results=str(code_results))}\n\n"
context_message = context_message.strip()
# Setup Prompt with Primer or Conversation History
messages = generate_chatml_messages_with_context(
conversation_primer,
user_query,
context_message=context_message,
conversation_log=conversation_log,
model_name=model,
max_prompt_size=max_prompt_size,

23
src/khoj/processor/conversation/google/gemini_chat.py
View File

@@ -157,9 +157,7 @@ def converse_gemini(
"""
# Initialize Variables
current_date = datetime.now()
compiled_references = "\n\n".join({f"# {item}" for item in references})
conversation_primer = prompts.query_prompt.format(query=user_query)
compiled_references = "\n\n".join({f"# File: {item['file']}\n## {item['compiled']}\n" for item in references})
if agent and agent.personality:
system_prompt = prompts.custom_personality.format(
@@ -191,20 +189,19 @@ def converse_gemini(
completion_func(chat_response=prompts.no_online_results_found.format())
return iter([prompts.no_online_results_found.format()])
if ConversationCommand.Code in conversation_commands and not is_none_or_empty(code_results):
conversation_primer = (
f"{prompts.code_executed_context.format(code_results=str(code_results))}\n{conversation_primer}"
)
if ConversationCommand.Online in conversation_commands or ConversationCommand.Webpage in conversation_commands:
conversation_primer = (
f"{prompts.online_search_conversation.format(online_results=str(online_results))}\n{conversation_primer}"
)
context_message = ""
if not is_none_or_empty(compiled_references):
conversation_primer = f"{prompts.notes_conversation.format(query=user_query, references=compiled_references)}\n\n{conversation_primer}"
context_message = f"{prompts.notes_conversation.format(query=user_query, references=compiled_references)}\n\n"
if ConversationCommand.Online in conversation_commands or ConversationCommand.Webpage in conversation_commands:
context_message += f"{prompts.online_search_conversation.format(online_results=str(online_results))}\n\n"
if ConversationCommand.Code in conversation_commands and not is_none_or_empty(code_results):
context_message += f"{prompts.code_executed_context.format(code_results=str(code_results))}\n\n"
context_message = context_message.strip()
# Setup Prompt with Primer or Conversation History
messages = generate_chatml_messages_with_context(
conversation_primer,
user_query,
context_message=context_message,
conversation_log=conversation_log,
model_name=model,
max_prompt_size=max_prompt_size,

27
src/khoj/processor/conversation/offline/chat_model.py
View File

@@ -158,7 +158,7 @@ def converse_offline(
# Initialize Variables
assert loaded_model is None or isinstance(loaded_model, Llama), "loaded_model must be of type Llama, if configured"
offline_chat_model = loaded_model or download_model(model, max_tokens=max_prompt_size)
compiled_references_message = "\n\n".join({f"{item['compiled']}" for item in references})
compiled_references = "\n\n".join({f"# File: {item['file']}\n## {item['compiled']}\n" for item in references})
tracer["chat_model"] = model
current_date = datetime.now()
@@ -176,8 +176,6 @@ def converse_offline(
day_of_week=current_date.strftime("%A"),
)
conversation_primer = prompts.query_prompt.format(query=user_query)
if location_data:
location_prompt = prompts.user_location.format(location=f"{location_data}")
system_prompt = f"{system_prompt}\n{location_prompt}"
@@ -187,31 +185,34 @@ def converse_offline(
system_prompt = f"{system_prompt}\n{user_name_prompt}"
# Get Conversation Primer appropriate to Conversation Type
if conversation_commands == [ConversationCommand.Notes] and is_none_or_empty(compiled_references_message):
if conversation_commands == [ConversationCommand.Notes] and is_none_or_empty(compiled_references):
return iter([prompts.no_notes_found.format()])
elif conversation_commands == [ConversationCommand.Online] and is_none_or_empty(online_results):
completion_func(chat_response=prompts.no_online_results_found.format())
return iter([prompts.no_online_results_found.format()])
if ConversationCommand.Code in conversation_commands and not is_none_or_empty(code_results):
conversation_primer = (
f"{prompts.code_executed_context.format(code_results=str(code_results))}\n{conversation_primer}"
)
if ConversationCommand.Online in conversation_commands:
context_message = ""
if not is_none_or_empty(compiled_references):
context_message = f"{prompts.notes_conversation_offline.format(references=compiled_references)}\n\n"
if ConversationCommand.Online in conversation_commands or ConversationCommand.Webpage in conversation_commands:
simplified_online_results = online_results.copy()
for result in online_results:
if online_results[result].get("webpages"):
simplified_online_results[result] = online_results[result]["webpages"]
conversation_primer = f"{prompts.online_search_conversation_offline.format(online_results=str(simplified_online_results))}\n{conversation_primer}"
if not is_none_or_empty(compiled_references_message):
conversation_primer = f"{prompts.notes_conversation_offline.format(references=compiled_references_message)}\n\n{conversation_primer}"
context_message += (
f"{prompts.online_search_conversation_offline.format(online_results=str(simplified_online_results))}\n\n"
)
if ConversationCommand.Code in conversation_commands and not is_none_or_empty(code_results):
context_message += f"{prompts.code_executed_context.format(code_results=str(code_results))}\n\n"
context_message = context_message.strip()
# Setup Prompt with Primer or Conversation History
messages = generate_chatml_messages_with_context(
conversation_primer,
user_query,
system_prompt,
conversation_log,
context_message=context_message,
model_name=model,
loaded_model=offline_chat_model,
max_prompt_size=max_prompt_size,

23
src/khoj/processor/conversation/openai/gpt.py
View File

@@ -155,9 +155,7 @@ def converse(
"""
# Initialize Variables
current_date = datetime.now()
compiled_references = "\n\n".join({f"# {item['compiled']}" for item in references})
conversation_primer = prompts.query_prompt.format(query=user_query)
compiled_references = "\n\n".join({f"# File: {item['file']}\n## {item['compiled']}\n" for item in references})
if agent and agent.personality:
system_prompt = prompts.custom_personality.format(
@@ -188,22 +186,21 @@ def converse(
completion_func(chat_response=prompts.no_online_results_found.format())
return iter([prompts.no_online_results_found.format()])
if not is_none_or_empty(code_results):
conversation_primer = (
f"{prompts.code_executed_context.format(code_results=str(code_results))}\n{conversation_primer}"
)
if not is_none_or_empty(online_results):
conversation_primer = (
f"{prompts.online_search_conversation.format(online_results=str(online_results))}\n{conversation_primer}"
)
context_message = ""
if not is_none_or_empty(compiled_references):
conversation_primer = f"{prompts.notes_conversation.format(query=user_query, references=compiled_references)}\n\n{conversation_primer}"
context_message = f"{prompts.notes_conversation.format(references=compiled_references)}\n\n"
if not is_none_or_empty(online_results):
context_message += f"{prompts.online_search_conversation.format(online_results=str(online_results))}\n\n"
if not is_none_or_empty(code_results):
context_message += f"{prompts.code_executed_context.format(code_results=str(code_results))}\n\n"
context_message = context_message.strip()
# Setup Prompt with Primer or Conversation History
messages = generate_chatml_messages_with_context(
conversation_primer,
user_query,
system_prompt,
conversation_log,
context_message=context_message,
model_name=model,
max_prompt_size=max_prompt_size,
tokenizer_name=tokenizer_name,

4
src/khoj/processor/conversation/prompts.py
View File

@@ -118,6 +118,7 @@ Use my personal notes and our past conversations to inform your response.
Ask crisp follow-up questions to get additional context, when a helpful response cannot be provided from the provided notes or past conversations.
User's Notes:
-----
{references}
""".strip()
)
@@ -127,6 +128,7 @@ notes_conversation_offline = PromptTemplate.from_template(
Use my personal notes and our past conversations to inform your response.
User's Notes:
-----
{references}
""".strip()
)
@@ -328,6 +330,7 @@ Use this up-to-date information from the internet to inform your response.
Ask crisp follow-up questions to get additional context, when a helpful response cannot be provided from the online data or past conversations.
Information from the internet:
-----
{online_results}
""".strip()
)
@@ -337,6 +340,7 @@ online_search_conversation_offline = PromptTemplate.from_template(
Use this up-to-date information from the internet to inform your response.
Information from the internet:
-----
{online_results}
""".strip()
)

32
src/khoj/processor/conversation/utils.py
View File

@@ -22,6 +22,7 @@ from transformers import AutoTokenizer
from khoj.database.adapters import ConversationAdapters, ais_user_subscribed
from khoj.database.models import ChatModelOptions, ClientApplication, KhojUser
from khoj.processor.conversation import prompts
from khoj.processor.conversation.offline.utils import download_model, infer_max_tokens
from khoj.search_filter.date_filter import DateFilter
from khoj.search_filter.file_filter import FileFilter
@@ -259,8 +260,9 @@ def generate_chatml_messages_with_context(
query_images=None,
vision_enabled=False,
model_type="",
context_message="",
):
"""Generate messages for ChatGPT with context from previous conversation"""
"""Generate chat messages with appropriate context from previous conversation to send to the chat model"""
# Set max prompt size from user config or based on pre-configured for model and machine specs
if not max_prompt_size:
if loaded_model:
@@ -274,21 +276,27 @@ def generate_chatml_messages_with_context(
# Extract Chat History for Context
chatml_messages: List[ChatMessage] = []
for chat in conversation_log.get("chat", []):
message_notes = f'\n\n Notes:\n{chat.get("context")}' if chat.get("context") else "\n"
message_context = ""
if chat["by"] == "khoj" and "excalidraw" in chat["intent"].get("type", ""):
message_context += chat.get("intent").get("inferred-queries")[0]
if not is_none_or_empty(chat.get("context")):
references = "\n\n".join(
{f"# File: {item['file']}\n## {item['compiled']}\n" for item in chat.get("context") or []}
)
message_context += f"{prompts.notes_conversation.format(references=references)}\n\n"
if not is_none_or_empty(chat.get("onlineContext")):
message_context += f"{prompts.online_search_conversation.format(online_results=chat.get('onlineContext'))}"
if not is_none_or_empty(message_context):
reconstructed_context_message = ChatMessage(content=message_context, role="user")
chatml_messages.insert(0, reconstructed_context_message)
role = "user" if chat["by"] == "you" else "assistant"
if chat["by"] == "khoj" and "excalidraw" in chat["intent"].get("type"):
message_content = chat.get("intent").get("inferred-queries")[0] + message_notes
else:
message_content = chat["message"] + message_notes
message_content = construct_structured_message(message_content, chat.get("images"), model_type, vision_enabled)
message_content = construct_structured_message(chat["message"], chat.get("images"), model_type, vision_enabled)
reconstructed_message = ChatMessage(content=message_content, role=role)
chatml_messages.insert(0, reconstructed_message)
if len(chatml_messages) >= 2 * lookback_turns:
if len(chatml_messages) >= 3 * lookback_turns:
break
messages = []
@@ -299,6 +307,8 @@ def generate_chatml_messages_with_context(
role="user",
)
)
if not is_none_or_empty(context_message):
messages.append(ChatMessage(content=context_message, role="user"))
if len(chatml_messages) > 0:
messages += chatml_messages
if not is_none_or_empty(system_message):

2
src/khoj/routers/api_chat.py
View File

@@ -120,7 +120,7 @@ def add_files_filter(request: Request, filter: FilesFilterRequest):
file_filters = ConversationAdapters.add_files_to_filter(request.user.object, conversation_id, files_filter)
return Response(content=json.dumps(file_filters), media_type="application/json", status_code=200)
except Exception as e:
logger.error(f"Error adding file filter {filter.filename}: {e}", exc_info=True)
logger.error(f"Error adding file filter {filter.filenames}: {e}", exc_info=True)
raise HTTPException(status_code=422, detail=str(e))