Test

.github/propmts/ai-developer-guide.prompt.yml

@@ -0,0 +1,23 @@
+name: AI Developer Guide - Apply Best Practices
+description: Based on the AI developer guide propose best practices to improve code, documentation, developer experience and more.
+model: gpt-4o-mini
+modelParameters:
+  temperature: 0.5
+messages:
+  - role: system
+    content: You are a tech lead and propose improvements to my project based on the AI developer gide.
+  - role: user
+    content: |
+      Read the developer guide at
+      https://github.com/dwmkerr/ai-developer-guide
+      You MUST follow the rules in this guide.
+      Propose improvements to my codebase.
+# testData:
+#   - input: |
+#       The quick brown fox jumped over the lazy dog.
+#       The dog was too tired to react.
+#     expected: Summary - A fox jumped over a lazy, unresponsive dog.
+# evaluators:
+#   - name: Output should start with 'Summary -'
+#     string:
+#       startsWith: 'Summary -'

CHANGELOG.md

@@ -1,5 +1,27 @@
 # Changelog
 
+## [0.15.0](https://github.com/dwmkerr/terminal-ai/compare/v0.14.0...v0.15.0) (2025-06-12)
+
+
+### Features
+
+* **devx:** ai developer guide integrated to GitHub AI ([#104](https://github.com/dwmkerr/terminal-ai/issues/104)) ([b2e127a](https://github.com/dwmkerr/terminal-ai/commit/b2e127a1701be188be49ad64f5d9bc5800e958d6))
+
+
+### Bug Fixes
+
+* **devx:** correct prompt path ([c1b06e0](https://github.com/dwmkerr/terminal-ai/commit/c1b06e0197c002b2ec762bff3752512f40d589f6))
+
+## [0.14.0](https://github.com/dwmkerr/terminal-ai/compare/v0.13.2...v0.14.0) (2025-05-24)
+
+
+### Features
+
+* **config:** improve provider/model setup ([#99](https://github.com/dwmkerr/terminal-ai/issues/99)) ([8cd60bf](https://github.com/dwmkerr/terminal-ai/commit/8cd60bfe709a3c810ea3f8badef8e0138d6b7c08))
+* **docs:** langfuse integration ([#96](https://github.com/dwmkerr/terminal-ai/issues/96)) ([2471b63](https://github.com/dwmkerr/terminal-ai/commit/2471b63fde611d32ae9e0ec07382f62bbc620bd0))
+* **docs:** msty config guide ([#97](https://github.com/dwmkerr/terminal-ai/issues/97)) ([1e0c9de](https://github.com/dwmkerr/terminal-ai/commit/1e0c9de3c6a59979416b80b7cbee48ed7460aa47))
+* **providers:** add LiteLLM integration ([#103](https://github.com/dwmkerr/terminal-ai/issues/103)) ([8dc9f8f](https://github.com/dwmkerr/terminal-ai/commit/8dc9f8ff6aae4ab47d31305b4a5bff4d22faad62))
+
 ## [0.13.2](https://github.com/dwmkerr/terminal-ai/compare/v0.13.1...v0.13.2) (2025-05-01)
 
 

README.md

@@ -1,7 +1,7 @@
 <p align="center">
   <h2 align="center"><code>🧠 terminal-ai</code></h2>
       <h3 align="center">Effortless AI in the shell. Maintain your flow and be more effective.</h3>
-      <h5 align="center">Supports OpenAI, Gemini, Claude, Ollama and many more</h5>
+      <h5 align="center">Supports OpenAI, Gemini, Claude, Ollama, MSTY, LiteLLM and many more.</h5>
   <p align="center">
       <img src="./docs/casts/terminal-ai-homepage.svg" alt="Demo Recording of Terminal AI" style="max-width: 100%;">
   </p>
@@ -12,7 +12,8 @@
     <a href="#commands">Commands</a> |
     <a href="#configuration">Configuration</a> |
     <a href="#api-key">API Key</a> |
-    <a href="#documentation">Documentation</a> |
+    <a href="#integrations">Integrations</a> |
+    <a href="#documentation">Documentation</a>
   </p>
   <p align="center">
     <a href="https://github.com/dwmkerr/terminal-ai/actions/workflows/cicd.yaml"><img src="https://github.com/dwmkerr/terminal-ai/actions/workflows/cicd.yaml/badge.svg" alt="cicd"></a>
@@ -333,9 +334,17 @@ export AI_MODEL="gpt-3.5-turbo"                 # Optional.
 
 To configure multiple providers or advanced options, check the [Configuration Documentation](./docs/configuration.md).
 
+## Integrations
+
+Integration is available out-of-the-box for [Langfuse](https://langfuse.com/). See the [Integrations Guide](./docs/integrations.md) for details.
+
 ## Documentation
 
 - [Configuration](./docs/configuration.md)
 - [Developer Guide](./docs/developer-guide.md)
+- [Integrations Guide](./docs/integrations.md)
+- [Providers: Setting Up Ollama](./docs/providers/ollama.md)
 - [Experimental Features](./docs/experimental-features.md)
 - [Providers: Setting Up Ollama](./docs/providers/ollama.md)
+- [Providers: Setting Up MSTY](./docs/providers/msty.md)
+- [Providers: Setting Up LiteLLM](./docs/providers/litellm.md)

docs/configuration.md

@@ -213,4 +213,24 @@ providers:
     baseURL: https://api.anthropic.com/v1
     model: claude-3-opus-20240229
     apiKey: '123'
+  msty:
+    name: msty
+    type: openai_compatible
+    baseURL: http://localhost:10000/v1/
+    model: gemma3:1b
+    apiKey: 'notused'
+  ollama:
+    name: ollama
+    type: openai_compatible
+    baseURL: http://localhost:11434/v1/
+    model: gemma3:1b
+    apiKey: 'notused'
+  litellm:
+    name: litellm
+    type: openai_compatible
+    baseURL: http://localhost:4000/
+    model: claude-3.7
+    apiKey: 'notused'
 ```
+
+Additional configuration options can be set for [Integrations](./integrations.md).

docs/integrations.md

@@ -0,0 +1,57 @@
+# Integrations
+
+Terminal AI can integrate with various external services to enhance its capabilities. This document explains the available integrations and how to configure them.
+
+<!-- vim-markdown-toc GFM -->
+
+- [Langfuse](#langfuse)
+    - [Setup](#setup)
+    - [Configuration Options](#configuration-options)
+
+<!-- vim-markdown-toc -->
+
+## Langfuse
+
+[Langfuse](https://langfuse.com) is an open-source observability platform for LLM applications. Integrating Terminal AI with Langfuse allows you to:
+
+- Track conversations and model interactions
+- Monitor latency, token usage and costs
+- Debug issues in production
+- Analyze model performance over time
+
+![Langfuse Integration Screenshot](./images/integrations-langfuse.png)
+
+### Setup
+
+To configure Langfuse integration with Terminal AI:
+
+1. Create a Langfuse account at [langfuse.com](https://langfuse.com) or self-host it
+2. Create a new Lanfuse project
+3. Obtain your Langfuse secret key and public key
+4. Add the Langfuse configuration to your `~/.ai/config.yaml` file:
+
+```yaml
+integrations:
+  langfuse:
+    secretKey: "your-langfuse-secret-key"
+    publicKey: "your-langfuse-public-key"
+    baseUrl: "https://cloud.langfuse.com"
+    traceName: "terminal-ai"
+```
+
+You can validate your Langfuse configuration by running:
+
+```bash
+ai check
+```
+
+### Configuration Options
+
+| Configuration Option | Default Value | Description |
+|----------------------|---------------|-------------|
+| `secretKey` | (empty) | Your Langfuse secret key (required) |
+| `publicKey` | (empty) | Your Langfuse public key (required) |
+| `baseUrl` | `https://cloud.langfuse.com` | The Langfuse API endpoint |
+| `traceName` | `terminal-ai` | The name to use for traces in Langfuse |
+
+If you're self-hosting Langfuse, change the `baseUrl` to your instance URL.

docs/providers/litellm.md

@@ -0,0 +1,107 @@
+# litellm
+
+[LiteLLM](https://www.litellm.ai/) can be used to proxy requests to many different inference providers, track LLM spend, manage keys and more:
+
+<img src="../images/litellm.png" width="600px" alt="LiteLLM Screenshot" />
+
+LiteLLM exposes providers using the [OpenAI API specification](https://docs.litellm.ai/docs/#call-100-llms-using-the-openai-inputoutput-format). This means that it can be used to proxy Terminal AI requests from OpenAI format to non-OpenAI compatible providers.
+
+As an example, we will configure LiteLLM to proxy OpenAI SDK requests to Anthropic Claude. Note that as Claude has an [OpenAI compatible endpoint](https://docs.anthropic.com/en/api/openai-sdk) you could simply configure Terminal AI to use this (see the [Configuration Guide](../configuration.md)), however this example shows how LiteLLM can translate requests to Claudes own API format.
+
+Create a [LiteLLM configuration file](https://docs.litellm.ai/docs/proxy/configs) with Claude 3.7 configured as a model:
+
+```bash
+cat << EOF > litellm_config.yaml
+model_list:
+  - model_name: claude-3.7
+    litellm_params:
+      model: claude-3-7-sonnet-20250219
+      api_key: "os.environ/ANTHROPIC_API_KEY"
+EOF
+```
+
+Set your Anthropic API key. You could also provide an API base URL if you would like to call a custom endpoint (such as an internal AI API gateway):
+
+```bash
+ANTHROPIC_API_KEY="***"
+ANTHROPIC_API_BASE="https://api.anthropic.com/v1"
+```
+
+Now run the proxy container:
+
+```bash
+docker run \
+    -v $(pwd)/litellm_config.yaml:/app/config.yaml \
+    -e ANTHROPIC_API_KEY \
+    -e ANTHROPIC_API_BASE \
+    -p 4000:4000 \
+    ghcr.io/berriai/litellm:main-latest \
+    --config /app/config.yaml --detailed_debug
+```
+
+Run a completion to confirm that your configuration is correct:
+
+```bash
+curl --location 'http://0.0.0.0:4000/chat/completions' \
+--header 'Content-Type: application/json' \
+--data '{
+      "model": "claude-3.7",
+      "messages": [
+        {
+          "role": "user",
+          "content": "what llm are you"
+        }
+      ]
+    }'
+# example output:
+# {
+#   "model":"claude-3-7-sonnet-20250219",
+#   "message":{
+#     "content":"I am Claude, an AI assistant created by Anthropic...",
+#     "role":"assistant"
+#     ...
+```
+
+Now run `ai init` to configure your provider, using the following details:
+
+- Provider Type: OpenAI Compatible
+- API Key: `notused` (LiteLLM has required keys loaded into its proxy)
+- Base URL: `http://localhost:4000/`
+- Model: `claude-3.7`
+- Provider name: `litellm`
+
+Choose 'yes' for 'Set as current provider' and 'Test API Key & Configuration'. You will see output similar to the below:
+
+```
+✔ Set as current provider? Yes
+✔ Test API Key & Configuration? Yes
+✔ Checking internet connection...
+✔ Checking Base URL http://localhost:4000/...
+✔ Checking API key...
+✔ Checking Model claude-3.7...
+✔ Checking API key rate limit...
+```
+
+At this point you will be able to interface with inference providers via the LiteLLM proxy:
+
+```
+✔ chat: bash one liner for a rainbow
+claude:
+
+    for i in {1..7}; do echo -e "\033[3${i}mRainbow\033[0m"; done
+
+Run this to see a simple rainbow text effect in your terminal.
+```
+
+You can also manually add the LiteLLM provider details to your [Configuration File](../configuration.md):
+
+```yaml
+provider: litellm # set the current provider
+providers:
+  litellm:
+    name: litellm
+    type: openai_compatible
+    baseURL: http://localhost:4000/
+    model: claude-3.7
+    apiKey: notused
+```

docs/providers/msty.md

@@ -4,7 +4,47 @@ Terminal AI works out of the box with Msty.
 
 To get started, first install: https://msty.app/
 
-Run Msty and configure a model. As an example we'll use Cohere as a remote provider:
+Once Msty is installed and running, you'll have access to various models through its local API server. Msty will automatically start an API server on port 10000 by default.
 
-- New Remote Model Provider: Cohere AI
-- API Key: (Provide your API key)
+Run Msty and configure a model. As an example we'll use Ollama as a local provider.
+
+Then run `ai init` to configure your provider, using the following details:
+
+- Provider Type: OpenAI Compatible
+- API Key: `notused` (by default, Msty doesn't need a key, but Terminal AI requires one to be configured)
+- Base URL: `http://localhost:10000/v1/`
+- Model: Choose from one of your available models (e.g. `gemma3:1b` or `llama3.2:latest`)
+
+Choose 'yes' for 'Set as current provider' and 'Test API Key & Configuration'. You will see output similar to the below:
+
+```
+✔ Set as current provider? Yes
+✔ Test API Key & Configuration? Yes
+✔ Checking internet connection...
+✔ Checking Base URL http://localhost:10000/v1/...
+✔ Checking API key...
+✔ Checking Model gemma3:1b...
+✔ Checking API key rate limit...
+```
+
+At this point you will be able to chat using your local Msty installation:
+
+```
+✔ chat: hi msty
+msty: Hello! How can I assist you today?
+```
+
+You can also manually add the Msty provider details to your [Configuration File](../configuration.md):
+
+```yaml
+provider: msty
+providers:
+  msty:
+    name: msty
+    type: openai_compatible
+    baseURL: http://localhost:10000/v1/
+    model: gemma3:1b
+    apiKey: notused
+```
+
+If you're using Msty with a remote provider (like Cohere), make sure that model is properly configured in your Msty app before attempting to use it with Terminal AI.

package.json

@@ -2,7 +2,7 @@
   "name": "@dwmkerr/terminal-ai",
   "description": "Effortless AI in your terminal.",
   "type": "commonjs",
-  "version": "0.13.2",
+  "version": "0.15.0",
   "main": "./dist/cli.js",
   "bin": {
     "ai": "dist/cli.js"

src/chat-pipeline/stages/ensure-api-key.ts

@@ -20,5 +20,5 @@ export async function ensureApiKey(executionContext: ExecutionContext) {
 
   //  Initialise, this will mutate execution context to set the key, or die
   //  trying.
-  return await init(executionContext, true);
+  return await init(executionContext);
 }

src/cli.ts

@@ -81,7 +81,7 @@ const cli = async (program: Command, executionContext: ExecutionContext) => {
     .command("init")
     .description("Set or update configuration")
     .action(async () => {
-      const nextCommand = await init(executionContext, true);
+      const nextCommand = await init(executionContext);
       //  The only possible next action is chat or quit.
       if (nextCommand === Commands.Chat) {
         return chat(

src/commands/init/init-first-run.ts

@@ -14,7 +14,6 @@ import { initSetProviderApiKey } from "./init-set-provider-api-key";
  */
 export async function initFirstRun(
   executionContext: ExecutionContext,
-  askNextAction: boolean,
 ): Promise<Commands> {
   const interactive = executionContext.isTTYstdin;
 
@@ -33,8 +32,6 @@ To get a free key follow the guide at:`,
   );
   await initSetProviderApiKey(executionContext);
 
-  //  We now continue with a regular init run (which'll offer the option to
-  //  update the model, whatever). However, we don't ask the user if they want
-  //  to change their API key.
-  return await initRegularRun(executionContext, false, askNextAction);
+  //  We now continue with a regular init run.
+  return await initRegularRun(executionContext);
 }