Commit graph

1685 commits

Author SHA1 Message Date
Debanjum Singh Solanky
8a9055f918 Restrict server messages show in echo area to main server files 2023-03-27 04:59:55 +07:00
Debanjum Singh Solanky
ae535a06eb Configure Khoj chat using khoj.el by setting OpenAI API key in Emacs 2023-03-27 04:59:54 +07:00
Debanjum Singh Solanky
36b17d4ae0 Generalize the directory from config extraction elisp method 2023-03-27 03:44:03 +07:00
Debanjum Singh Solanky
924424c754 Throw actionable exceptions when content types or chat not configured 2023-03-27 02:47:44 +07:00
Debanjum Singh Solanky
359a2cacef Fix khoj--server-running to work with unconfigured or external server
- If khoj server started outside emacs, khoj--server-ready should be set
to true by khoj--server-running method (instead of waiting for proc msg)

- If khoj server is unconfigured the /config/types endpoint wouldn't
return anything. Using config/data/default allows checking khoj server
running status without requiring it to be configured as well
2023-03-27 02:45:59 +07:00
Debanjum Singh Solanky
d7fb9a596e Auto configure server before loading khoj-menu
If the config hasn't changed there'll be no update. If config has
changed indexing will get triggered asynchronously. But user cannot
make query till indexing done

As easier to know when server ready to configure
2023-03-27 02:44:02 +07:00
Debanjum Singh Solanky
8a21aff438 Make khoj.el server start, stop, restart, setup methods interactive
No need to erase temporary buffers before working on them
2023-03-27 01:53:15 +07:00
Debanjum Singh Solanky
cb40a96c85 Index configured org files from khoj.el
- Set `khoj-org-files-index' to list of files to index
- Defaults to indexing org-agenda-files
- Uses khoj server api to configure org files to index
2023-03-27 01:05:26 +07:00
Debanjum Singh Solanky
50760acc37 Wait for Khoj server to get ready before opening khoj.el transient menu
- Use process filter, sentinel to mark when khoj server is ready or not
- Display server messages for visibility into server boot-up process
- Wait until server ready to open khoj transient menu in Emacs
  Until then khoj features wouldn't work anyway, so avoids confusion
2023-03-26 13:00:01 +07:00
Debanjum Singh Solanky
82eb4bfd0d Setup Khoj server on opening khoj from with Emacs
- Create helper methods to check, stop, restart, setup khoj server
- (Ask to) setup khoj server on calling khoj main entrypoint function
2023-03-26 10:12:06 +07:00
Debanjum Singh Solanky
99d19dcf43 Start Khoj server from Emacs using khoj.el 2023-03-26 09:38:46 +07:00
Debanjum Singh Solanky
c92d79118a Install Khoj server from Emacs using khoj.el 2023-03-26 08:50:03 +07:00
Debanjum Singh Solanky
e281a498b4 Style Khoj search org buffer via elisp instead of in-buffer settings 2023-03-26 06:34:18 +07:00
Debanjum Singh Solanky
4f655d20ae Style Khoj chat directly via elisp instead of via in-buffer settings 2023-03-26 06:03:30 +07:00
Debanjum Singh Solanky
f6ff7b1beb Render foonote reference links as superscript for Khoj Chat on Emacs 2023-03-26 05:33:08 +07:00
Debanjum Singh Solanky
285a2b86d2 Use aiohttp version 3.8.4 as 4.x breaks docker image build 2023-03-26 05:33:02 +07:00
Debanjum Singh Solanky
67c850a4ac Add retry logic to OpenAI API queries to increase Chat tenacity
- Move completion and chat_completion into helper methods under utils.py
- Add retry with exponential backoff on OpenAI exceptions using
  tenacity package. This is officially suggested and used by other
  popular GPT based libraries
2023-03-26 05:12:35 +07:00
Debanjum
0aebf624fc
Improve Khoj Chat in Emacs, Server
Merge pull request #192 from debanjum/improvements-to-khoj-chat-in-emacs

### Khoj Chat on Emacs Improvements
- d78454d Load Khoj Chat buffer before asking for query to provide context
- 93e2aff Use org footnotes to add references, allows jump to def on click
- 5e9558d Stylize reference links as superscripts and show definition on hover
- bc71c19 Use `m` or `C-x m` in-buffer keybindings to send messages to Khoj

### Khoj Chat Server Improvements
- 27217a3 Time chat API sub-components for performance analysis
- 508b217 Update Chat API, Logs, Interfaces to store, use references as list
- d4b3866 Truncate message logs to below max supported prompt size by chat model
- cf28f10 Register separate timestamps for user query and response by Khoj Chat
2023-03-25 05:49:27 +07:00
Debanjum Singh Solanky
ff846f05c5 Clean-up khoj.el based on linting helpers and manual review 2023-03-25 05:47:49 +07:00
Debanjum Singh Solanky
7e36f421f9 Truncate message logs to below max supported prompt size by model
- Use tiktoken to count tokens for chat models
- Make conversation turns to add to prompt configurable via method
  argument to generate_chatml_messages_with_context method
2023-03-25 05:13:56 +07:00
Debanjum Singh Solanky
4725416fbd Use shortcut keybindings in buffer to ease sending messages to Khoj 2023-03-25 05:06:01 +07:00
Debanjum Singh Solanky
508b2176b7 Update Chat API, Logs, Interfaces to store, use references as list
- Remove the need to split by magic string in emacs and chat interfaces
- Move compiling references into string as context for GPT to GPT layer
- Update setup in tests to use new style of setting references
- Name first argument to converse as more appropriate "references"
2023-03-24 22:10:11 +07:00
Debanjum Singh Solanky
b08745b541 Keep chat messages at 1 empty line visible distance in khoj.el
- Clean redundant concat, format string
- Improve variable name to emojified sender
2023-03-24 22:10:11 +07:00
Debanjum Singh Solanky
27217a330d Time chat API sub-components for performance analysis
Time and the search query extraction, search and response generation
components
2023-03-24 20:39:41 +07:00
Debanjum Singh Solanky
5e9558d39d Stylize references shown as footnote links in chat messages
- Render references as superscript
- Show reference definitions on hover over reference links to ease access
- Truncate reference def shown on hover to 70 char
  - Add continuation suffix, ..., when reference definition truncated
2023-03-24 20:38:05 +07:00
Debanjum Singh Solanky
cf28f104c7 Register separate timestamps for user query and response by Khoj Chat 2023-03-24 18:31:58 +07:00
Debanjum Singh Solanky
93e2aff786 Add references as org footnotes instead of links 2023-03-24 18:31:42 +07:00
Debanjum Singh Solanky
d78454d4ad Load Khoj Chat buffer before asking for query to provide context 2023-03-24 13:43:46 +07:00
Debanjum
4070d13a96
Create Khoj Chat Interface in Emacs
Merge pull request #191 from debanjum/create-chat-interface-on-emacs

- Render conversation history in a read-only org-mode buffer for Khoj Chat
- Add `chat` as a transient action in the Khoj transient menu
- Style chat messages as org-mode entries
  - Put received date in property drawer and keep it hidden/folded by default
  - Add Khoj chat response as child entry of the users associated question org entry
    This allows folding back-n-forth between user and Khoj for easier viewing
  - Render source notes snippets used as references for response as org-mode links
    Hovering mouse on link or opening links shows reference note snippets used
2023-03-22 16:32:40 -06:00
Debanjum Singh Solanky
863933daaa Resolve build issues found by melpazoid 2023-03-23 02:25:34 +04:00
Debanjum Singh Solanky
e9ca04af0d Require dash, org to run ERT tests for khoj.el 2023-03-23 01:46:26 +04:00
Debanjum Singh Solanky
06df394d6c Style chat messages as org-mode entries in Emacs
- Style Message as Org Entries instead of List
- Put khoj response as child of user query entry
  - Improves color coding for readability
  - Allows folding each back-n-forth
- Put timestamp of message received into property drawer
- Use standardized time format for new and old chat messages
2023-03-22 12:00:43 -06:00
Debanjum Singh Solanky
364e6c11af Render chat history from API in chat buffer on first run
- Generalize the render-chat-response method to handle rendering
  history or chat response from chat API reponse

- Trigger rendering of khoj chat history if Khoj chat buffer not
  created for this session yet
2023-03-22 12:00:35 -06:00
Debanjum Singh Solanky
36b52fdd0a Properly escape reference links before rendering
- Use org-insert-link method to improve link rendering robustness
  Previous simple mechanism to crete org-links would result in links
  escaping out of formating. Use a user-facing org-mode method to
  remove/reduce probability of this

- Replace newlines with space to render reference notes as links
2023-03-22 11:05:38 -06:00
Debanjum Singh Solanky
72f63a6ef7 Add basic chat interface for Khoj on Emacs
- Query khoj chat API to get Khoj Chat response to user message
- Render chat messages as a org-mode list in format:
  - [sender-name]: *[message]*
    - /[receive-date]/
- Add references as org links with context visible on hover,
  but no jump to note
- Require dash library for khoj.el to simplify list manipulation.
  Use `-map-indexed' method from dash
2023-03-22 10:47:55 -06:00
Debanjum Singh Solanky
e4d67694e1 Add search to method, variable names meant for khoj search in khoj.el
In preparation to introduce Khoj chat in Emacs
2023-03-21 21:44:11 -06:00
Debanjum Singh Solanky
98e5ea4940 Fix name of default encoder to replace in multi-lingual model setup docs 2023-03-21 20:38:17 -06:00
Debanjum Singh Solanky
2f6284872d Mention Khoj needs Python version 3.10 or lower in docs 2023-03-20 15:18:19 -06:00
Debanjum Singh Solanky
a9b81975f2 Fix encoder model name to configure multilingual search in Readme
See comment in issue #98 for stale model name comment
2023-03-19 17:27:53 -06:00
Debanjum
b351cfb8a0
Add Search Actor to Improve Querying Notes for Khoj Chat
Merge pull request #189 from debanjum/add-search-actor-to-improve-notes-lookup-for-chat

### Introduce Search Actor
Search actor infers Search Queries from user's message
- Capabilities
  - Use previous messages to add context to current search queries[^1]
    This improves quality of responses in multi-turn conversations. 
  - Deconstruct users message into multiple search queries to lookup notes[^2]
  - Use relative date awareness to add date filters to search queries[^3]

- Chat Director now does the following:
  1. [*NEW*] Use Search Actor to generate search queries from user's message
  2. Retrieve relevant notes from Knowledge Base using the Search queries
  3. Pass retrieved relevant notes to Chat Actor to respond to user

### Add Chat Quality Tests 
- Test Search Actor capabilities
- Mark Chat Director Tests for Relative Date, Multiple Search Queries as Expected Pass

### Give More Search Results as Context to Chat Actor
- Loosen search results score threshold to work better for searches with date filters
- Pass more search results (up to 5 from 2) as context to Chat Actor to improve inference

[^1]: Multi-Turn Example
Q: "When did I go to Mars?"
Search: "When did I go to Mars?"
A: "You went to Mars in the future"
Q: "How was that experience?"
Search: "How my Mars experience?"
*This gives better context for the Chat actor to respond* 
[^2]: Deconstruct Example: 
Is Alpha older than Beta? => What is Alpha's age? & When was Beta born?

[^3]: Date Example: 
Convert user messages containing relative dates like last month, yesterday to date filters on specific dates like dt>="2023-03-01"
2023-03-18 18:02:12 -06:00
Debanjum Singh Solanky
601ff2541b Revert to using GPT to extract search queries from users message
- Reasons:
  - GPT can extract date aware search queries with date filters
    better than ChatGPT given the same prompt.
  - Need quality more than cost savings for now.
  - Need to figure ways to improve prompt for ChatGPT before using it
2023-03-18 17:56:13 -06:00
Debanjum Singh Solanky
e28526bbc9 Extract search queries from users message using ChatGPT as Search Actor
- Reasons
  - ChatGPT should be better at following instructions than GPT
  - At 1/10th the cost, it's much cheaper than using older GPT models
2023-03-18 16:33:24 -06:00
Debanjum Singh Solanky
939d7731da Fix-up Search Actor GPT's response for decoding it as valid JSON 2023-03-18 16:30:55 -06:00
Debanjum Singh Solanky
f63fd0995e Pass more search results as context to Chat Actor to improve inference 2023-03-18 16:30:55 -06:00
Debanjum Singh Solanky
10836dedee Search should return user message if GPT response is not valid JSON
Previously would throw if GPT response is not valid JSON. Better to
return original message to use for search instead
2023-03-18 16:30:55 -06:00
Debanjum Singh Solanky
08f5fb315f Add answers to context for Search Actor to generate relevant queries
Update Search Actor prompt with answers, more precise primer and
two more examples for context

Mark the 3 chat quality tests using answer as context to generate
queries as expected to pass. Verify that the 3 tests pass now, unlike
before when the Search Actor did not have the answers for context
2023-03-18 16:30:55 -06:00
Debanjum Singh Solanky
f09bdd515b Expect Chat Director can extract relative dates using new Search Actor 2023-03-18 16:30:55 -06:00
Debanjum Singh Solanky
36c7389b46 Test Search Actor generating search query from Chat History 2023-03-18 16:30:55 -06:00
Debanjum Singh Solanky
2600cc9d4d Test Search Actor extracting relative dates & multiple questions 2023-03-18 16:30:55 -06:00
Debanjum Singh Solanky
45cb510421 Loosen search results score thresold used by chat for more context 2023-03-18 16:30:55 -06:00