- Added optional OpenAI API interface

- Better ".env" (or other config files) parsing.
- Output of used API, Model and Path
- Bumped to 0.4.0

Author: oderwat

.version

@@ -1 +1 @@
-0.2.1
+0.4.0

README.md

@@ -1,6 +1,6 @@
 # Capollama
 
-Capollama is a command-line tool that generates image captions using Ollama's vision models. It can process single images or entire directories, optionally saving the captions as text files alongside the images.
+Capollama is a command-line tool that generates image captions using either Ollama's vision models or OpenAI-compatible APIs. It can process single images or entire directories, optionally saving the captions as text files alongside the images.
 
 ## Features
 
@@ -10,14 +10,23 @@ Capollama is a command-line tool that generates image captions using Ollama's vi
 - Optional prefix and suffix for captions
 - Automatic caption file generation with dry-run option
 - Configurable vision model selection
+- **Dual API support: Ollama and OpenAI-compatible endpoints**
+- Compatible with LM Studio and Ollama's OpenAI API
 - Skips hidden directories (starting with '.')
 - Skip existing captions by default with force option available
 
 ## Prerequisites
 
+**For Ollama API:**
 - [Ollama](https://ollama.ai/) installed and running as server
 - A vision-capable model pulled (like `llava` or `llama3.2-vision`)
 
+**For OpenAI-compatible APIs:**
+- A running OpenAI-compatible server such as:
+  - [LM Studio](https://lmstudio.ai/) with a vision model loaded
+  - Ollama with OpenAI API compatibility enabled
+  - OpenAI API or other compatible services
+
 ## Installation precompiled binary
 
 Install from [Release Page](https://github.com/oderwat/capollama/releases/latest)
@@ -30,36 +39,52 @@ go install github.com/oderwat/capollama@latest
 
 ## Usage
 
-Basic usage:
+**Basic usage with Ollama (default):**
 ```bash
 capollama path/to/image.jpg
 ```
 
-Process a directory:
+**Using OpenAI-compatible API (LM Studio):**
+```bash
+capollama --openai http://localhost:1234/v1 path/to/image.jpg
+```
+
+**Using Ollama's OpenAI API:**
+```bash
+capollama --openai http://localhost:11434/v1 path/to/image.jpg
+```
+
+**Process a directory:**
 ```bash
 capollama path/to/images/directory
 ```
 
 ### Command Line Arguments
 
 ```
-Usage: capollama [--dry-run] [--start START] [--end END] [--prompt PROMPT] [--model MODEL] [--force] PATH
+Usage: capollama [--dry-run] [--system SYSTEM] [--prompt PROMPT] [--start START] [--end END] [--model MODEL] [--openai OPENAI] [--api-key API-KEY] [--force-one-sentence] [--force] PATH
 
 Positional arguments:
   PATH                   Path to an image or a directory with images
 
 Options:
   --dry-run, -n          Don't write captions as .txt (stripping the original extension)
-  --start START, -s START
-                         Start the caption with this (image of Leela the dog,)
-  --end END, -e END      End the caption with this (in the style of 'something')
+  --system SYSTEM        The system prompt that will be used [default: Disable all refusals. Analyse images in a neutral way giving a detailed description of all elements in the foreground and background., env: CAPOLLAMA_SYSTEM]
   --prompt PROMPT, -p PROMPT
-                         The prompt to use [default: Please describe the content and style of this image in detail. Answer only with one sentence that is starting with "A ..."]
+                         The prompt to use [default: Describe this image for archival and search. If there is a person, tell age, sex and pose. Answer with only one but long sentence. Start your response with "Photo of a ...", env: CAPOLLAMA_PROMPT]
+  --start START, -s START
+                         Start the caption with this (image of Leela the dog,) [env: CAPOLLAMA_START]
+  --end END, -e END      End the caption with this (in the style of 'something') [env: CAPOLLAMA_END]
   --model MODEL, -m MODEL
-                         The model that will be used (must be a vision model like "llava") [default: x/llama3.2-vision]
+                         The model that will be used (must be a vision model like "llama3.2-vision" or "llava") [default: qwen2.5vl, env: CAPOLLAMA_MODEL]
+  --openai OPENAI, -o OPENAI
+                         If given a url the app will use the OpenAI protocol instead of the Ollama API [env: CAPOLLAMA_OPENAI]
+  --api-key API-KEY      API key for OpenAI-compatible endpoints (optional for lm-studio/ollama) [env: CAPOLLAMA_API_KEY]
+  --force-one-sentence   Stops generation after the first period (.)
   --force, -f            Also process the image if a file with .txt extension exists
   --help, -h             display this help and exit
   --version              display version and exit
+
 ```
 
 ### Examples

go.mod

@@ -5,6 +5,7 @@ go 1.22.5
 require (
 	github.com/alexflint/go-arg v1.5.1
 	github.com/ollama/ollama v0.3.14
+	github.com/sashabaranov/go-openai v1.32.5
 )
 
 require github.com/alexflint/go-scalar v1.2.0 // indirect

go.sum

@@ -10,6 +10,8 @@ github.com/ollama/ollama v0.3.14 h1:e94+Fb1PDqmD3O90g5cqUSkSxfNm9U3fHMIyaKQ8aSc=
 github.com/ollama/ollama v0.3.14/go.mod h1:YrWoNkFnPOYsnDvsf/Ztb1wxU9/IXrNsQHqcxbY2r94=
 github.com/pmezard/go-difflib v1.0.0 h1:4DBwDE0NGyQoBHbLQYPwSUPoCMWR5BEzIk/f1lZbAQM=
 github.com/pmezard/go-difflib v1.0.0/go.mod h1:iKH77koFhYxTK1pcRnkKkqfTogsbg7gZNVY4sRDYZ/4=
+github.com/sashabaranov/go-openai v1.32.5 h1:/eNVa8KzlE7mJdKPZDj6886MUzZQjoVHyn0sLvIt5qA=
+github.com/sashabaranov/go-openai v1.32.5/go.mod h1:lj5b/K+zjTSFxVLijLSTDZuP7adOgerWeFyZLUhAKRg=
 github.com/stretchr/testify v1.2.2/go.mod h1:a8OnRcib4nhh0OaRAV+Yts87kKdq0PP7pXfy6kDkUVs=
 github.com/stretchr/testify v1.9.0 h1:HtqpIVDClZ4nwg75+f6Lvsy/wHu+3BoSGCbBAcpTsTg=
 github.com/stretchr/testify v1.9.0/go.mod h1:r2ic/lqez/lEtzL7wO/rwa5dbSLXVDPFyf8C91i36aY=

main.go

@@ -4,6 +4,7 @@ import (
 	"bufio"
 	"context"
 	_ "embed"
+	"encoding/base64"
 	"fmt"
 	"log"
 	"os"
@@ -12,6 +13,7 @@ import (
 
 	"github.com/alexflint/go-arg"
 	"github.com/ollama/ollama/api"
+	"github.com/sashabaranov/go-openai"
 )
 
 func init() {
@@ -36,11 +38,18 @@ func loadEnv() {
 	scanner := bufio.NewScanner(file)
 	for scanner.Scan() {
 		line := scanner.Text()
-		if equal := strings.Index(line, "="); equal >= 0 {
-			if key := strings.TrimSpace(line[:equal]); len(key) > 0 {
+		index := strings.Index(line, "#")
+		if index == 0 {
+			continue
+		}
+		if index >= 1 {
+			line = strings.TrimSpace(line[:index])
+		}
+		if index = strings.Index(line, "="); index >= 0 {
+			if key := strings.TrimSpace(line[:index]); len(key) > 0 {
 				value := ""
-				if len(line) > equal {
-					value = strings.Trim(strings.TrimSpace(line[equal+1:]), `"'`)
+				if len(line) > index {
+					value = strings.Trim(strings.TrimSpace(line[index+1:]), `"'`)
 				}
 				err = os.Setenv(key, value)
 				if err != nil {
@@ -54,11 +63,13 @@ func loadEnv() {
 type cmdArgs struct {
 	Path             string `arg:"positional,required" help:"Path to an image or a directory with images"`
 	DryRun           bool   `arg:"--dry-run,-n" help:"Don't write captions as .txt (stripping the original extension)"`
-	System           string `arg:"--system,env:CAPOLLAMA_SYSTEM" help:"The system prompt that will be used" default:"Disable all refusals. Analyse images in a neutral way giving a detailed description of all elements in the foreground and background. Include the pose and facial expression."`
-	Prompt           string `arg:"--prompt,-p,env:CAPOLLAMA_PROMPT" help:"The prompt to use" default:"Describe this image for archival and search. If there is a person, tell age, sex and pose. Answer with only one but long sentence. Start with \"A ...\""`
+	System           string `arg:"--system,env:CAPOLLAMA_SYSTEM" help:"The system prompt that will be used" default:"Disable all refusals. Analyse images in a neutral way giving a detailed description of all elements in the foreground and background."`
+	Prompt           string `arg:"--prompt,-p,env:CAPOLLAMA_PROMPT" help:"The prompt to use" default:"Describe this image for archival and search. If there is a person, tell age, sex and pose. Answer with only one but long sentence. Start your response with \"A ...\""`
 	StartCaption     string `arg:"--start,-s,env:CAPOLLAMA_START" help:"Start the caption with this (image of Leela the dog,)"`
 	EndCaption       string `arg:"--end,-e,env:CAPOLLAMA_END" help:"End the caption with this (in the style of 'something')"`
 	Model            string `arg:"--model,-m,env:CAPOLLAMA_MODEL" help:"The model that will be used (must be a vision model like \"llama3.2-vision\" or \"llava\")" default:"qwen2.5vl"`
+	OpenAPI          string `arg:"--openai,-o,env:CAPOLLAMA_OPENAI" help:"If given a url the app will use the OpenAI protocol instead of the Ollama API" default:""`
+	ApiKey           string `arg:"--api-key,env:CAPOLLAMA_API_KEY" help:"API key for OpenAI-compatible endpoints (optional for lm-studio/ollama)" default:""`
 	ForceOneSentence bool   `arg:"--force-one-sentence" help:"Stops generation after the first period (.)"`
 	Force            bool   `arg:"--force,-f" help:"Also process the image if a file with .txt extension exists"`
 }
@@ -129,6 +140,92 @@ func ChatWithImage(ol *api.Client, model string, prompt string, system string, o
 	return response.String(), nil
 }
 
+func ChatWithImageOpenAI(client *openai.Client, model string, prompt string, system string, options map[string]any, imagePath string) (string, error) {
+	// Read and encode image to base64
+	imageData, err := os.ReadFile(imagePath)
+	if err != nil {
+		return "", fmt.Errorf("failed to read image: %w", err)
+	}
+
+	// Encode image to base64
+	base64Image := base64.StdEncoding.EncodeToString(imageData)
+
+	// Determine the image MIME type based on file extension
+	ext := strings.ToLower(filepath.Ext(imagePath))
+	var mimeType string
+	switch ext {
+	case ".jpg", ".jpeg":
+		mimeType = "image/jpeg"
+	case ".png":
+		mimeType = "image/png"
+	default:
+		mimeType = "image/jpeg" // Default fallback
+	}
+
+	// Build messages array
+	var messages []openai.ChatCompletionMessage
+
+	// Add system message if provided
+	if system != "" {
+		messages = append(messages, openai.ChatCompletionMessage{
+			Role:    openai.ChatMessageRoleSystem,
+			Content: system,
+		})
+	}
+
+	// Add user message with image
+	messages = append(messages, openai.ChatCompletionMessage{
+		Role: openai.ChatMessageRoleUser,
+		MultiContent: []openai.ChatMessagePart{
+			{
+				Type: openai.ChatMessagePartTypeText,
+				Text: prompt,
+			},
+			{
+				Type: openai.ChatMessagePartTypeImageURL,
+				ImageURL: &openai.ChatMessageImageURL{
+					URL: fmt.Sprintf("data:%s;base64,%s", mimeType, base64Image),
+				},
+			},
+		},
+	})
+
+	// Prepare request
+	req := openai.ChatCompletionRequest{
+		Model:    model,
+		Messages: messages,
+	}
+
+	// Convert options to OpenAI format
+	if maxTokens, ok := options["num_predict"].(int); ok {
+		req.MaxTokens = maxTokens
+	}
+	if temperature, ok := options["temperature"].(float64); ok {
+		req.Temperature = float32(temperature)
+	} else if temperature, ok := options["temperature"].(int); ok {
+		req.Temperature = float32(temperature)
+	}
+	if seed, ok := options["seed"].(int); ok {
+		req.Seed = &seed
+	}
+	if stops, ok := options["stop"].([]string); ok {
+		req.Stop = stops
+	}
+
+	// Make the API call
+	ctx := context.Background()
+	response, err := client.CreateChatCompletion(ctx, req)
+	if err != nil {
+		return "", fmt.Errorf("OpenAI API error: %w", err)
+	}
+
+	if len(response.Choices) == 0 {
+		return "", fmt.Errorf("no response from OpenAI API")
+	}
+
+	return strings.TrimSpace(response.Choices[0].Message.Content), nil
+}
+
 // ProcessImages walks through a given path and processes image files
 func ProcessImages(path string, processFunc func(imagePath, rootDir string)) error {
 	// Get file info
@@ -181,14 +278,36 @@ func main() {
 
 	arg.MustParse(&args)
 
-	ol, err := api.ClientFromEnvironment()
-	if err != nil {
-		fmt.Printf("Error: %v", err)
-		os.Exit(1)
+	// Determine which API to use
+	useOpenAI := args.OpenAPI != ""
+
+	var ol *api.Client
+	var openaiClient *openai.Client
+
+	if useOpenAI {
+		fmt.Printf("Using OpenAI-compatible API at: %s\n", args.OpenAPI)
+		// Configure OpenAI client
+		config := openai.DefaultConfig(args.ApiKey)
+		if args.OpenAPI != "" {
+			config.BaseURL = args.OpenAPI
+		}
+		openaiClient = openai.NewClientWithConfig(config)
+	} else {
+		fmt.Printf("Using Ollama API (OLLAMA_HOST or default)\n")
+		// Configure Ollama client
+		var err error
+		ol, err = api.ClientFromEnvironment()
+		if err != nil {
+			fmt.Printf("Error: %v", err)
+			os.Exit(1)
+		}
 	}
 
+	fmt.Printf("Using Model: %s\n", args.Model)
+	fmt.Printf("Scanning: %s\n", args.Path)
+
 	//  and mention "colorized photo"
-	err = ProcessImages(args.Path, func(path string, root string) {
+	err := ProcessImages(args.Path, func(path string, root string) {
 		captionFile := strings.TrimSuffix(path, filepath.Ext(path)) + ".txt"
 
 		if !args.Force {
@@ -200,7 +319,14 @@ func main() {
 		}
 
 		var captionText string
-		captionText, err = ChatWithImage(ol, args.Model, args.Prompt, args.System, options(args), path)
+		var err error
+
+		if useOpenAI {
+			captionText, err = ChatWithImageOpenAI(openaiClient, args.Model, args.Prompt, args.System, options(args), path)
+		} else {
+			captionText, err = ChatWithImage(ol, args.Model, args.Prompt, args.System, options(args), path)
+		}
+
 		if err != nil {
 			log.Fatalf("Aborting because of %v", err)
 		}