Merge branch 'main' into doc_updates

kagisearch · Nov 7, 2024 · 6a96aa6 · 6a96aa6
2 parents 0af5f79 + c5eca2a
commit 6a96aa6
Show file tree

Hide file tree

Showing 11 changed files with 191 additions and 115 deletions.
diff --git a/docs/common/company/hiring-kagi.md b/docs/common/company/hiring-kagi.md
@@ -5,12 +5,10 @@ next:
 ---
 # Jobs at Kagi Search
 
-**We receive a high volume of applications, which means our response times can be slow.**
-
 We review every application, but this takes time.
 To increase your chances of a timely response:
 
-- Put real effort into your application
+- Put effort like you mean it into your application
 - Clearly demonstrate why you're a strong fit for the role
 - Be detailed, but get to the point. Quality over quantity.
 

diff --git a/docs/kagi/ai/assistant.md b/docs/kagi/ai/assistant.md
@@ -4,16 +4,17 @@
 
 <video src="./media/assistant.mp4" width="720" type="video/mp4" autoplay muted loop playsinline disablepictureinpicture />
 
-The Assistant by Kagi combines top large language models (LLM) with (optional) results from Kagi Search, making it the perfect companion for creative, research, and programming tasks — alongside everything else you can think of! And all included in a single subscription!
+The Assistant by Kagi combines top large language models (LLM) with optional results from Kagi Search, making it the perfect companion for creative, research, and programming tasks — alongside everything else you can think of! All this is included in a single subscription!
 
+>**NOTE:** The Assistant is currently available to users with the [Kagi Ultimate plan](../plans/ultimate-plan.md).
 
 ## Features
 
-- Access to the latest and most performant large language models from OpenAI, Anthropic, Meta, Google and Mistral
+- Access to the latest and most performant large language models from OpenAI, Anthropic, Meta, Google, and Mistral
 - Multiple [custom assistants](#custom-assistants)
-- The ability to control whether the Assistant has web access (powered by Kagi Search).
+- The ability to control whether the Assistant has web access (powered by Kagi Search)
 - Applying Kagi Search [Lenses](../features/lenses.md) and [Personalized Results](../features/website-info-personalized-results.md) to the Assistant searches
-- Switch between multiple threads from a single window
+- Switch between multiple threads within a single window
 - Saving Assistant threads
 - Uploading files to use as context
 - Altering the Assistant configuration within the thread
@@ -27,37 +28,37 @@ The Assistant by Kagi combines top large language models (LLM) with (optional) r
 
 ## Privacy
 
-When you use the Assistant by Kagi, your data is never used to train AI models, and no account information is shared with the LLM providers. By default, threads expire after 24 hours of inactivity (can be adjusted in [settings](https://kagi.com/settings/?p=assistant)).
+When you use the Assistant by Kagi, your data is never used to train AI models, and no account information is shared with the LLM providers. By default, threads expire after 24 hours of inactivity. This behavior can be adjusted in the [settings](https://kagi.com/settings/?p=assistant).
 
 ## Using the Assistant
 
-The Assistant can be accessed via the Control Center located in the top right corner of all Kagi pages or by clicking on the link in the top right corner of the Kagi homepage. [The direct link](https://kagi.com/assistant) to the Assistant can also be used. 
+The Assistant can be accessed via the Control Center located in the top right corner of all Kagi pages or by clicking on the link in the top right corner of the Kagi homepage. [The direct link](https://kagi.com/assistant) to the Assistant can also be used.
 
 When you first access the Assistant, you will be greeted by a familiar-looking landing page, allowing you to get right into using it.
 You can either type your prompt or use voice input by pressing the microphone symbol.
-You can choose which LLM you wish to use by opening the dropdown menu just left of the prompt field.
+You can choose which LLM you wish to use by opening the dropdown menu just below the prompt field.
 
-Assistant's web access can be toggled via the button below the prompt field.
+The Assistant's web access can be toggled via the button below the prompt field.
 
-## What model to choose
+## Which model to choose
 
-We maintain [Kagi LLM benchmarking Project](./llm-benchmark.md) to measure the model quality in various scenarios.
+We maintain the [Kagi LLM Benchmarking Project](./llm-benchmark.md) to measure model quality in various scenarios.
 
 ## Threads
 
 The Assistant supports threads, allowing you to keep your bagel topping ideas separate from your weekend projects.
 
 The search bar enables you to search for that one elusive thread.
-Currently the search only looks through your thread titles.
+Currently, the search only looks through your thread titles.
 
 By default threads are kept for 24 hours after the last message.
-If keeping threads alive permanently better fits your workflow, this setting can be adjusted in [Assistant Settings](https://kagi.com/settings/?p=assistant).
+If keeping threads alive permanently better fits your workflow, you can adjust this setting in [Assistant Settings](https://kagi.com/settings/?p=assistant).
 Please note that the thread saving setting is applied **when the thread is created**.
 
 Threads can be renamed, added to Quick Access, shared, deleted, and downloaded via the `⋮` button which is displayed when you hover over the thread.
 
 The Quick Access feature can be used to better organize your threads but it will not influence the thread saving setting.
-In other words, threads added to Quick Access will also be deleted after 24 hours unless you have chosen otherwise in the settings.
+In other words, threads added to Quick Access will also be deleted after 24 hours, unless you have chosen otherwise in the settings.
 
 You can also **preserve** threads with the star symbol, located to the left of the thread title.
 Preserved threads do not expire.
@@ -86,7 +87,7 @@ Important considerations for file uploads:
 **Processing time:** Larger files may take a few moments to process.
 **Context retention:** Uploaded file content remains in the conversation context for subsequent messages.
 
-The Assistant supports various file formats across different categories:
+The Assistant supports various file formats across different categories, including:
 
 | File Type | Supported Formats |
 |:--|:--|
@@ -145,15 +146,15 @@ The Assistant provides access to the following LLMs:
 	- Claude 3.5 Sonnet
 	- Claude 3 Opus
 - OpenAI
-	- GPT 3.5 Turbo
 	- GPT 4
 	- GPT 4 Turbo
 	- GPT 4o
+	- GPT 4o Mini
 - Mistral AI
 	- Mistral Pixtral
 	- Mistral Large
 - Google
-	- Gemini Pro
+	- Gemini 1.5 Pro
 - Meta
 	- Llama 3.1 405B
 
@@ -166,6 +167,18 @@ You can quickly access Assistant from Kagi Search by using the following [bangs]
 
 You can also access the **Code** Custom Assistant with the `!code` bang.
 
+## URL Parameters
+
+You can specify a particular model in the Assistant's URL by including a `profile` parameter.
+`https://kagi.com/assistant?profile=gpt-4o&q=%s`
+The available model names include: `gpt-4`, `gpt-4-turbo`, `gpt-4o`, `gpt-4o-mini`, `gemini-pro`, `mistral-nemo`, `mistral-large`, `claude-3-haiku`, `claude-3-sonnet`, `claude-3-opus`, `llama-3-405b`, `code`.
+
+This can also be used with custom assistants, as described on the [custom assistant documentation](./custom-assistants.md#url-parameters).
+
+The `internet` parameter can be used to turn on and off internet access, set to `true` to enable, anything else to disable. This overrides the internet setting of the profile used.
+
+The `lens` parameter can be used to set the lens if internet access is enabled. The value of this is the lowercase format of the lens name, for example, `https://kagi.com/assistant?lens=programming&q=%s` will use the Programming lens.
+
 ## Availability
 
 The Assistant is available to all Kagi Ultimate members.
@@ -175,8 +188,5 @@ We also offer an Ultimate upgrade for Family Plans, you can upgrade from the [Fa
 
 ## FAQ
 
-**Q**: Does the Assistant have file upload capability?
-**A**: The Assistant will have file upload capabilities very soon (work in progress). In the meantime, you can continue to use the [legacy Assistant](https://kagi.com/v1_assistant).
-
-**Q**: What is Kagi’s stance about using LLMs in search?
-**A**: We are continue to relentlessly focus on the core search experience and build thoughtfully integrated features on top of it. Read more about it in our [AI Integration Philosophy](../why-kagi/ai-philosophy.md) page.
+**Q**: What is Kagi’s stance about using LLMs in search?\
+**A**: We continue to relentlessly focus on the core search experience and build thoughtfully integrated features on top of it. Read more about it in our [AI Integration Philosophy](../why-kagi/ai-philosophy.md) page.
diff --git a/docs/kagi/ai/custom-assistants.md b/docs/kagi/ai/custom-assistants.md
@@ -20,59 +20,6 @@
 	- Context Window
 - **Response Instructions**: Define how your assistant should interact (1500 characters max)
 
-## Using Custom Assistants
-
-Select your Custom Assistant from the model-selection dropdown menu below the prompt bar to begin interaction.
-
-## Example Configurations
-
-Here are a few example configurations you can use to get started. You can also visit [Anthropic’s Prompt Library](https://docs.anthropic.com/en/prompt-library/library) for more useful prompts.
-
-### Professional Writer Assistant
-```markdown
-Name: Writing Pro
-Bang: !write
-Internet Access: Enabled (for current writing trends and examples)
-Lens: Entire Internet
-Instructions: Respond in a professional tone with clear, concise language. Focus on grammar, style, and clarity. Provide constructive feedback on writing samples. When appropriate, suggest improvements while maintaining the original message's intent.
-```
-
-### Coding Mentor
-```markdown
-Name: Code Guide
-Bang: !code
-Internet Access: Enabled (for documentation and best practices)
-Lens: Programming
-Instructions: Explain programming concepts clearly with practical examples. Use best practices and modern conventions. Include comments in code examples. Focus on readability and maintainability. Highlight potential pitfalls and optimization opportunities.
-```
-
-### Academic Researcher
-```markdown
-Name: Research Aid
-Bang: !research
-Internet Access: Enabled (for current research and papers)
-Lens: Academic
-Instructions: Provide comprehensive, academic-style responses with clear citations. Focus on peer-reviewed sources when available. Present balanced viewpoints on complex topics. Highlight key findings and methodologies.
-```
-
-### Recipe Helper
-```markdown
-Name: Chef Assistant
-Bang: !chef
-Internet Access: Enabled (for recipes and cooking techniques)
-Lens: Recipes
-Instructions: Provide clear, step-by-step cooking instructions. Focus on practical cooking tips, ingredient substitutions, and technique explanations. Include serving sizes and preparation times.
-```
-
-### News Analyst
-```markdown
-Name: News Guide
-Bang: !news
-Internet Access: Enabled (for current events)
-Lens: World News
-Instructions: Analyze current events with balanced perspective. Provide context for news developments. Focus on factual reporting and multiple viewpoints. Summarize key points clearly.
-```
-
 ## Tips for Effective Instructions
 
 - Be specific about desired tone and style
@@ -95,3 +42,83 @@ Instructions: Analyze current events with balanced perspective. Provide context
 5. Update settings as your needs change
 
 Remember that well-crafted instructions lead to more accurate and useful responses. Take time to think about your specific needs and how to communicate them effectively.
+
+## Using Custom Assistants
+
+Select your Custom Assistant from the model-selection dropdown menu below the prompt bar to begin interaction.
+
+## URL Parameters
+
+You can specify a custom assistant by adding a `profile` parameter to the URL, using either the name (in lowercase format) or the UUID of your custom assistant.
+
+For instance, if you want to use a specific custom assistant, locate its UUID by navigating to your [Custom Assistants settings page](https://kagi.com/settings?p=assistant), and copy the long UUID found in the URL when editing the assistant. You can then create a custom bang template, like:
+
+```
+/assistant?q=%s&profile=c03f3098-9ead-408f-93f0-407a77e697db
+```
+
+This setup will direct the request to your specified assistant.
+
+More information about assistant URL parameters can be found in the [assistant documentation](./assistant.md#url-parameters).
+
+## Default Custom Assistants
+
+The Assistant comes with carefully pre-configured assistants designed to enhance your productivity.
+
+### Code Assistant
+Powered by the best AI model for programming tasks and equipped with web access, the Code Assistant excels at:
+
+- Writing efficient, clean code
+- Debugging and troubleshooting
+- Providing practical coding solutions
+
+Simply use the ```!code``` shortcut or select "Code" from the model dropdown to access this specialized assistant. Whether you're debugging a tricky issue, learning a new programming language, or seeking to optimize your code, the Code Assistant is your go-to programming companion.
+
+## Example Custom Configurations
+
+Here are a few example configurations you can use to get started. You can also visit [Anthropic’s Prompt Library](https://docs.anthropic.com/en/prompt-library/library) for more useful prompts.
+
+### Professional Writer Assistant
+| Setting | Value |
+|---------|-------|
+| Name | Writing Pro |
+| Bang | !write |
+| Internet Access | Enabled (for current writing trends and examples) |
+| Lens | Entire Internet |
+| Instructions | Respond in a professional tone with clear, concise language. Focus on grammar, style, and clarity. Provide constructive feedback on writing samples. When appropriate, suggest improvements while maintaining the original message's intent. |
+
+### Coding Mentor
+| Setting | Value |
+|---------|-------|
+| Name | Code Guide |
+| Bang | !code |
+| Internet Access | Enabled (for documentation and best practices) |
+| Lens | Programming |
+| Instructions | Explain programming concepts clearly with practical examples. Use best practices and modern conventions. Include comments in code examples. Focus on readability and maintainability. Highlight potential pitfalls and optimization opportunities. |
+
+### Academic Researcher
+| Setting | Value |
+|---------|-------|
+| Name | Research Aid |
+| Bang | !research |
+| Internet Access | Enabled (for current research and papers) |
+| Lens | Academic |
+| Instructions | Provide comprehensive, academic-style responses with clear citations. Focus on peer-reviewed sources when available. Present balanced viewpoints on complex topics. Highlight key findings and methodologies. |
+
+### Recipe Helper
+| Setting | Value |
+|---------|-------|
+| Name | Chef Assistant |
+| Bang | !chef |
+| Internet Access | Enabled (for recipes and cooking techniques) |
+| Lens | Recipes |
+| Instructions | Provide clear, step-by-step cooking instructions. Focus on practical cooking tips, ingredient substitutions, and technique explanations. Include serving sizes and preparation times. |
+
+### News Analyst
+| Setting | Value |
+|---------|-------|
+| Name | News Guide |
+| Bang | !news |
+| Internet Access | Enabled (for current events) |
+| Lens | World News |
+| Instructions | Analyze current events with balanced perspective. Provide context for news developments. Focus on factual reporting and multiple viewpoints. Summarize key points clearly. |
diff --git a/docs/kagi/ai/llm-benchmark.md b/docs/kagi/ai/llm-benchmark.md
@@ -6,36 +6,23 @@ Introducing the Kagi LLM Benchmarking Project, which evaluates major large langu
 
 The Kagi LLM Benchmarking Project uses an unpolluted benchmark to assess contemporary large language models (LLMs) through diverse, challenging tasks. Unlike standard benchmarks, our tests frequently change and are mostly novel, providing a rigorous evaluation of the models' capabilities, (hopefully) outside of what models saw in the training data to avoid benchmark overfitting. 
 
-Last updated **Oct 24, 2024**.
-
-
-| Model | Accuracy (%)| Tokens | Total Cost ($) | Median Latency (s) | Speed (tokens/sec) |
-|------------------------------------------|----------|--------|----------------|--------------------|--------------------|
-| OpenAI **gpt-4o** | 52 | 5889 | 0.11922 | 1.08 | 50.49 |
-| Together **meta-llama/Meta-Llama-3.1-405B-Instruct-Turbo** | 50 | 7767  | 0.07136  | 2.00    | 46.49 |  
-| OpenAI **o1-mini** | 50 | 29820| 0.37716 | 4.35 | n/a |
-| Anthropic **claude-3.5-sonnet-202410220** | 48 | 6595 | 0.12018 | 2.54 | 48.90 |
-| OpenAI **o1-preview** | 48 | 38440| 2.40306 | 9.29 | n/a |
-| OpenRouter **nvidia/llama-3.1-nemotron-70b-instruct** | 44 | 11989 | 0.00700 | 5.71 | 26.81 |
-| Mistral **large-latest** | 44 | 5097 | 0.06787 | 3.08 | 18.03 |
-| OpenAI **gpt-4o-mini** | 42 | 6029 | 0.00451 | 1.64 | 36.92 |
-| Groq **llama-3.1-70b-versatile** | 40 | 5190 | 0.00781 | 0.71 | 81.62 |
-| OpenRouter **x-ai/grok-2** | 40 | 6917 | 0.10141 | 2.47 | 41.15 |
-| OpenRouter **nousresearch/hermes-3-llama-3.1-405b:free** | 40 | 6075 | 0.00000 | 3.93 | 19.05 |
-| Reka **reka-core** | 36 | 6966 | 0.12401 | 6.21 | 17.56 |
-| DeepSeek **deepseek-chat** | 32 | 7310 | 0.00304 | 4.81 | 17.20 |
-| Anthropic **claude-3-haiku-20240307** | 28 | 5642 | 0.00881 | 1.33 | 55.46 |
-| Groq **llama-3.1-8b-instant** | 28 | 6628 | 0.00085 | 2.26 | 82.02 |
-| OpenRouter **mistralai/ministral-8b** | 28 | 5415 | 0.00120 | 1.12 | 72.76 |
-| DeepSeek **deepseek-coder** | 28 | 8079 | 0.00327 | 4.13 | 16.72 |
-| OpenAI **gpt-4** | 26 | 2477 | 0.33408 | 1.32 | 16.68 |
-| Mistral **open-mistral-nemo** | 22 | 4135 | 0.00323 | 0.65 | 82.65 |
-| Groq **gemma2-9b-it** | 22 | 4889 | 0.00249 | 1.69 | 54.39 |
-| OpenAI **gpt-3.5-turbo** | 22 | 1569 | 0.01552 | 0.51 | 45.03 |
-| Reka **reka-edge** | 20 | 5377 | 0.00798 | 2.02 | 46.87 |
-| Reka **reka-flash** | 16 | 5738 | 0.01668 | 3.28 | 28.75 |
-| GoogleGenAI **gemini-1.5-pro-exp-0801** | 14 | 4942 | 0.26325 | 1.82 | 28.19 |
-| GoogleGenAI **gemini-1.5-flash** | 14 | 5287 | 0.02777 | 3.02 | 21.16 |
+Last updated **Nov 4, 2024**.
+
+| Model | Accuracy (%) | Tokens | Total Cost ($) | Median Latency (s) | Speed (tokens/sec) |
+|-----------------------------|---------------|--------|----------------|------------------|-------------------|
+| **Anthropic** Claude-3.5-sonnet-20241022 | 43.33 | 9633 | 0.16649 | 2.38 | 55.53 |
+| **OpenAI** o1-mini | 41.67 | 42874 | 0.53454 | 4.62 | n/a |
+| **Anthropic** Claude-3.5-haiku-20241022 | 40.00 | 9738 | 0.05605 | 1.84 | 66.92 |
+| **OpenAI** gpt-4o | 40.00 | 11172 | 0.12801 | 1.59 | 49.90 |
+| **OpenRouter** nvidia/llama-3.1-nemotron-70b-instruct | 38.33 | 17413 | 0.00951 | 10.52 | 18.35 |
+| **OpenRouter** x-ai/grok-beta | 35.59 | 12187 | 0.21690 | 3.29 | 43.34 |
+| **Together** Meta-Llama-3.1-405B-Instruct-Turbo | 35.00 | 13306 | 0.10075 | 2.45 | 40.44 |
+| **OpenAI** gpt-4-turbo | 30.00 | 11460 | 0.40902 | 3.17 | 34.01 |
+| **Groq** llama-3.1-70b-versatile | 30.00 | 10879 | 0.01344 | 1.85 | 80.75 |
+| **OpenAI** gpt-4o-mini | 20.00 | 13181 | 0.00884 | 1.58 | 82.08 |
+| **OpenRouter** meta-llama/llama-3.2-3b-instruct | 18.33 | 13003 | 0.00081 | 0.89 | 94.52 |
+| **OpenAI** gpt-3.5-turbo | 8.33 | 4136 | 0.02132 | 0.72 | 59.10 |
+| **Anthropic** Claude-3-haiku-20240307 | 8.33 | 10387 | 0.01481 | 1.24 | 118.84 |
 
 
 The table includes metrics such as overall mode quality (measured as percent of correct responses), total tokens output (some models are less verbose by default, affecting both cost and speed), total cost to run the test, median response latency and average speed in tokens per second at the time of testing.
@@ -82,6 +69,7 @@ The table below is updated to the best of our abilities, feel free to submit cha
 | **GPT-4 (32k)**                    | 32K            | 60                    | 120                    |
 | **GPT-3.5-Turbo**                  | 16K            | 0.5                   | 1.5                    |
 | **Claude 3 Haiku**                 | 200K           | 0.25                  | 1.25                   |
+| Claude 3.5 Haiku                   | 200K           | 1                     | 5                      |
 | **Claude 3.5 Sonnet**              | 200K           | 3                     | 15                     |
 | **Claude 3 Opus**                  | 200K           | 15                    | 75                     |
 | **Gemini 1.5 Pro** (128K/1M)       | 1M             | 3.50/7                | 10.50/21               |

diff --git a/docs/kagi/ai/media/kagi_translate.png b/docs/kagi/ai/media/kagi_translate.png