Commit Graph

Select branches

Hide Pull Requests

master

Mono Color

• 2091044db5 Prefer agent chat model to extract document search queries Debanjum 2025-08-27 13:36:15 -07:00
• 7a42042488 Share context builder for chat final response across model types Debanjum 2025-08-26 23:32:17 -07:00
• 02e220f5f5 Pass args to context builder funcs grouped consistently Debanjum 2025-08-27 13:30:01 -07:00
• 4976b244a4 Set fast, deep think models for intermediary steps via admin panel Debanjum 2025-08-26 14:44:51 -07:00
• a99eb841ff Do not search documents when default tool selected by agent Debanjum 2025-08-25 00:40:18 -07:00
• a52a06ad9d Prefer olostep over firecrawl for webpage read by default Debanjum 2025-08-26 17:59:33 -07:00
• e150dc5a91 Improve copying message with math, file links to clipboard on web app Debanjum 2025-08-25 12:36:15 -07:00
• 15d1f39d0b Improve instruction to ai model for writing math expressions in LaTeX Debanjum 2025-08-25 12:35:08 -07:00
• 05dbb6a7c1 Drop unused generated_files arg from chat context Debanjum 2025-08-25 14:02:01 -07:00
• 8a16f5a2af Reduce logical complexity of constructing context from chat history Debanjum 2025-08-25 00:33:19 -07:00
• 1e81b51abc Support generating images with different aspect ratios Debanjum 2025-08-26 19:20:47 -07:00
• 5a2cae3756 Improve, simplify image generation prompts and context flow Debanjum 2025-08-25 00:27:17 -07:00
• 0fb6020f30 Remove model type check to construct structured messages Debanjum 2025-08-25 14:13:26 -07:00
• 386a17371d Fix identifying deepseek r1 model to process its thinking tokens Debanjum 2025-08-26 17:14:44 -07:00
• ff004d31ef Fix extracting inferred queries from chat history db Debanjum 2025-08-25 00:28:36 -07:00
• 892e4d4077 Fix system prompt construction for gemini models Debanjum 2025-08-24 18:22:31 -07:00
• 00c5aec614 Release Khoj version 2.0.0-beta.14 Debanjum 2025-08-23 12:12:33 -07:00
• b99ccbc4c3 Improve table styling, fix chat sidebar height on web app Debanjum 2025-08-23 01:30:47 -07:00
• 29ae476a26 Use groq with service tier auto to fallback to flex on rate limit Debanjum 2025-08-23 00:54:31 -07:00
• c89c5c7b46 Reorder ai model api columns on admin panel for readability Debanjum 2025-08-22 22:37:10 -07:00
• 464c1546b7 Support deepseek v3.1 via official deepseek api Debanjum 2025-08-22 22:13:15 -07:00
• 40488b3b68 Remove redundant exception for retry calls to gemini api Debanjum 2025-08-22 19:00:35 -07:00
• 8aa9c0f534 Reduce max reasoning tokens for gemini models Debanjum 2025-08-22 18:57:38 -07:00
• 2823c84bb4 Default to gemini 2.5 model series on init and for eval Debanjum 2025-08-22 18:44:26 -07:00
• c53a70c997 Share debug logs from github eval run for debugging Debanjum 2025-08-22 18:24:55 -07:00
• e2f377c27b Render file reference as link with file preview on hover/click in web app Debanjum 2025-08-21 00:06:21 -07:00
• d8b7e9c8a5 Handle unset content type key when indexing knowledge base on server Debanjum 2025-08-22 17:27:30 -07:00
• 3c3205bb06 Fix and improve file read, write handling in Obsidian Debanjum 2025-08-20 20:20:12 -07:00
• 48ed7afab8 Do not show <SYSTEM> instructions in chat session title on obsidian Debanjum 2025-08-20 19:17:27 -07:00
• 82dc7b115b Fix to allow khoj to delete content in obsidian write mode Debanjum 2025-08-20 16:40:05 -07:00
• 7645cbea3b Do not throw error when no edit blocks in write mode on obsidian Debanjum 2025-08-20 16:18:38 -07:00
• 2e6928c582 Limit retries to fix invalid edit blocks in obsidian write mode Debanjum 2025-08-20 17:25:09 -07:00
• c5e2373d73 Make khoj obsidian keyboard shortcuts toggle voice chat, chat history Debanjum 2025-08-20 16:14:43 -07:00
• d8b2df4107 Only show 3 recent files as context in obsidian file read, write mode Debanjum 2025-08-20 15:09:01 -07:00
• eb2f0ec6bc Persist open file access mode setting across restarts in obsidian Debanjum 2025-08-20 12:46:05 -07:00
• 2884853c98 Make plugin object accessible to chat, find similar panes in obsidian Debanjum 2025-08-20 12:44:14 -07:00
• 9f6aa922a2 Improve Khoj research tools, gpt-oss support and ai api usage Debanjum 2025-08-20 20:06:18 -07:00
• 13d26ae8b8 Wrap long words in train of thought shown on web app Debanjum 2025-08-20 12:32:24 -07:00
• fb0347a388 Truncate long words in stdout, stderr for context efficiency Debanjum 2025-08-19 23:31:45 -07:00
• dbc3330610 Tune reasoning effort, temp, top_p for gpt-oss models Debanjum 2025-08-19 23:29:42 -07:00
• 83d725d2d8 Extract thoughts of openai style models like gpt-oss from api response Debanjum 2025-08-19 23:26:45 -07:00
• f483a626b8 Simplify view file tool. Limit viewing upto 50 lines at a time Debanjum 2025-08-19 16:37:32 -07:00
• f5a4d106d1 Use instruction instead of query as code tool argument Debanjum 2025-08-19 00:04:29 -07:00
• c5a9c81479 Update khoj personality prompts with better style, capability guide Debanjum 2025-08-18 17:06:14 -07:00
• 2c91edbb25 Improve view file, code tool prompts. Format other research tool prompts Debanjum 2025-08-18 17:05:13 -07:00
• 452c794e93 Make regex search tool results look more like grep results Debanjum 2025-08-18 16:52:59 -07:00
• 9a8c707f84 Do not overwrite charts created in previous code tool use during research Debanjum 2025-08-18 20:19:29 -07:00
• e0007a31bb Update web UX when server side error or hit stop + no task running Debanjum 2025-08-18 23:26:37 -07:00
• 222cc19b7f Use subscriber type specific context window to generate response Debanjum 2025-08-18 23:24:00 -07:00
• ff73d30106 Fix max thinking budget for gemini models to generate final response Debanjum 2025-08-18 23:25:11 -07:00
• 34dca8e114 Fix passing temp kwarg to non-streaming openai completion endpoint Debanjum 2025-08-19 15:10:19 -07:00
• 8862394c15 Handle unset reasoning, response chunk from openai api while streaming Debanjum 2025-08-19 15:01:56 -07:00
• 14b4d4b663 Fix using non-reasoning openai model via responses API Debanjum 2025-08-18 21:34:04 -07:00
• e504141c07 Fix to calculate usage from openai api streaming completion Debanjum 2025-08-18 18:53:30 -07:00
• 573c6a32e1 Fix to create chat with custom agents from obsidian (#1216) Debanjum 2025-08-20 19:07:16 -07:00
• 4728098cad Fix to set agent for new chat created from obsidian Debanjum 2025-08-21 07:33:25 +05:30
• a2a3eb8be6 Update chat_view.ts Fh26697 2025-08-20 17:01:47 +02:00
• b3015f6837 Update chat_view.ts Fh26697 2025-08-20 16:58:08 +02:00
• 916534226a Chats are not using specified Agent Fh26697 2025-08-19 15:49:41 +02:00
• fa143d45b9 Fix passing images to official openai models using the responses api Debanjum 2025-08-17 16:30:43 -07:00
• a494a766a4 Fix eval github workflow and show more logs to debug its startup Debanjum 2025-08-15 16:04:50 -07:00
• 25e549d683 Show connection lost toast if disconnect while processing chat request Debanjum 2025-08-15 14:17:01 -07:00
• 59bfaf9698 Fix to indicate ws disconnect on web app & save interrupted research Debanjum 2025-08-15 12:48:28 -07:00
• 3eb8cce984 Retry if hit gemini rate limit. Return friendly message if retries fail Debanjum 2025-08-14 17:57:41 -07:00
• 4274f58dbd Show more specific warning to llm on duplicate tool use during research Debanjum 2025-08-15 14:17:35 -07:00
• caf0b994e8 Fix handling failure to select default chat tools Debanjum 2025-08-14 18:04:08 -07:00
• 7251b25c66 Handle null reference exceptions when rendering files context Debanjum 2025-08-15 15:59:57 -07:00
• 20347e21c2 Reduce noisy indexing logs Debanjum 2025-08-12 12:06:43 -07:00
• bd82626084 Release Khoj version 2.0.0-beta.13 Debanjum 2025-08-11 22:29:06 -07:00
• cbeefb7f94 Update researcher prompt to handle ambiguous queries. Clear stale text Debanjum 2025-08-11 22:15:44 -07:00
• 0a6d87067d Fix to have researcher let the coder tool write code Debanjum 2025-08-11 22:04:43 -07:00
• 0186403891 Limit retry to transient openai API errors. Return non-empty tool output Debanjum 2025-08-11 21:53:21 -07:00
• 41f89cf7f3 Handle price, responses of models served via Groq Debanjum 2025-08-11 19:31:33 -07:00
• b2d26088dc Use openai responses api to interact with official openai models Debanjum 2025-08-09 00:10:34 -07:00
• 564adb24a7 Add support for GPT 5 model series Debanjum 2025-08-08 12:35:55 -07:00
• 0e1615acc8 Fix grep files tool to work with line start, end anchors Debanjum 2025-08-08 15:20:28 -07:00
• a79025ee93 Limit max queries allowed per doc search tool call. Improve prompt Debanjum 2025-08-08 12:25:16 -07:00
• a3bb7100b4 Speed up app development using a faster, modern toolchain (#1196) Debanjum 2025-08-09 12:27:20 -07:00
• 80cce7b439 Fix server, web app to reuse prebuilt deps on dev container setup Debanjum 2025-08-01 18:31:25 -07:00
• 0a0b97446c Avoid `click' v8.2.2 server dependency as it breaks pypi validation Debanjum 2025-08-01 23:35:56 -07:00
• f2bd07044e Speed up github workflows by not installing cuda server dependencies Debanjum 2025-08-01 23:23:44 -07:00
• 8ad38dfe11 Switch to Bun instead of Deno (or Yarn) for faster web app builds Debanjum 2025-07-31 23:12:55 -07:00
• b86430227c Dedupe and move dev dependencies out from web app production builds Debanjum 2025-07-31 23:59:32 -07:00
• 791ebe3a97 Format web app code with prettier recommendations Debanjum 2025-08-01 00:25:53 -07:00
• c8e07e86e4 Format server code with ruff recommendations Debanjum 2025-08-01 00:10:34 -07:00
• 4a3ed9e5a4 Replace isort, black with ruff for faster linting, formatting Debanjum 2025-07-16 13:51:14 -07:00
• 8700fb8937 Use UV, Deno for faster setup of development container Debanjum 2025-06-08 19:07:38 -07:00
• d2940de367 Use Deno for speed, package locks in dev setup, github workflows Debanjum 2025-06-08 19:07:06 -07:00
• 006b958071 Use UV to install server for speed, package locks in dev setup, workflows Debanjum 2025-06-08 19:01:01 -07:00
• e0f363d718 Use UV to manage python version, env on khoj computer Debanjum 2025-06-08 16:30:52 -07:00
• 0387b86a27 Use portable comparator to get flags used to call dev_setup.sh Debanjum 2025-06-03 16:22:06 -07:00
• c6670e815a Drop Server Side Indexer, Native Offline Chat, Old Migration Scripts (#1212) Debanjum 2025-07-31 20:26:08 -05:00
• 892d57314e Update test setup to index test data after old indexing code removed Debanjum 2025-07-11 14:35:05 -07:00
• d9d24dd638 Drop old code to sync files on server filesystem. Clean cli, init paths Debanjum 2025-07-03 15:27:06 -07:00
• b1f2737c9a Drop native offline chat support with llama-cpp-python Debanjum 2025-07-03 01:49:18 -07:00
• 3f8cc71aca Drop old pre 1.0 khoj config migration scripts Debanjum 2025-07-03 15:04:55 -07:00
• 9096f628d0 Release Khoj version 2.0.0-beta.12 Debanjum 2025-07-31 18:13:17 -07:00
• a6923fac76 Improve description of query arg to semantic, web search tool Debanjum 2025-07-31 17:22:53 -07:00
• 2e13c9a007 Buffer thought chunks on server side for more performant ws streaming Debanjum 2025-07-31 17:21:06 -07:00
• fba4ad27f7 Extract thought stream from reasoning_content of openai model providers Debanjum 2025-07-31 16:00:59 -07:00