chore: cleanup

pyproject.toml

@@ -43,7 +43,7 @@ license = "MIT"
 [project.optional-dependencies]
 cli = ["rich", "typer"]
 extras = ["black", "pygments", "humanize", "natsort"]
-llm = ["litellm", "python-dotenv", "pillow"]
+llm = ["llmling-agent[litellm,pillow]"]
 icons = ["pyconify"]
 yaml = ["yamling"]
 httpx = ["httpx"]

src/jinjarope/llmfilters.py

@@ -6,7 +6,7 @@
 from dotenv import load_dotenv
 
 # import litellm
-from jinjarope import htmlfilters, lazylitellm
+from jinjarope import lazylitellm
 
 
 load_dotenv()
@@ -15,52 +15,6 @@
 litellm = lazylitellm.LazyLiteLLM()
 
 
-def llm_complete(
-    prompt: str,
-    system_prompt: str | None = None,
-    context: str | None = None,
-    model: str | None = None,
-    token: str | None = None,
-    base_url: str | None = None,
-    **kwargs: Any,
-) -> str:
-    """Complete a prompt using the LLM API.
-
-    Args:
-        prompt: The prompt to complete.
-        system_prompt: The system prompt to set context for the model.
-        context: Additional context for the prompt.
-        model: The model to use.
-        token: The API token.
-        base_url: The base URL of the API.
-        kwargs: Additional keyword arguments passed to litellm.completion.
-
-    Returns:
-        The completed text from the LLM.
-
-    Raises:
-        ValueError: If the API response is invalid or missing expected data.
-    """
-    messages: list[dict[str, str]] = []
-    if system_prompt:
-        messages.append({"role": "system", "content": system_prompt})
-    if context:
-        messages.append({"role": "user", "content": context})
-    messages.append({"role": "user", "content": prompt})
-
-    response = litellm.completion(
-        model=model or os.getenv("OPENAI_MODEL", ""),
-        api_key=token or os.getenv("OPENAI_API_TOKEN"),
-        api_base=base_url or os.getenv("OPENAI_API_BASE"),
-        messages=messages,
-        **kwargs,
-    )
-    if not response.choices or not response.choices[0].message:
-        msg = "Invalid API response: missing choices or message"
-        raise ValueError(msg)
-    return response.choices[0].message.content or ""
-
-
 def llm_generate_image(
     prompt: str,
     model: str | None = None,
@@ -103,73 +57,9 @@ def llm_generate_image(
     return None
 
 
-def llm_analyze_image(
-    image_url: str,
-    prompt: str | None = None,
-    model: str = "gpt-4-vision-preview",
-    token: str | None = None,
-    base_url: str | None = None,
-    encode_b64: bool = False,
-    **kwargs: Any,
-) -> str:
-    """Analyze an image using an LLM vision model and return the analysis as a string.
-
-    Args:
-        image_url: The URL of the image to analyze.
-        prompt: A prompt to guide the image analysis. If None, use a default prompt.
-        model: The name of the model to use. Defaults to "gpt-4-vision-preview".
-        token: The API token (key) for authentication.
-               If None, it will use the OPENAI_API_KEY environment variable.
-        base_url: The base URL for the API endpoint.
-                  If None, the default URL for the model will be used.
-        encode_b64: Whether to encode the image to base64 before sending it to the API.
-                    (required for some models)
-        kwargs: Additional keyword arguments passed to litellm.completion.
-
-    Returns:
-        The analysis of the image as a string.
-
-    Raises:
-        ValueError: If the image_url is empty or invalid.
-        requests.RequestException: If there's an error downloading the image.
-        Exception: If there's an error in making the API call or processing the response.
-    """
-    if not image_url or not image_url.strip():
-        msg = "Image URL cannot be empty"
-        raise ValueError(msg)
-
-    prompt = prompt or "Analyze this image and describe what you see in detail."
-    image_str = htmlfilters.url_to_b64(image_url) if encode_b64 else image_url
-    completion_kwargs: dict[str, Any] = {
-        "model": model,
-        "messages": [
-            {
-                "role": "user",
-                "content": [
-                    {"type": "text", "text": prompt},
-                    {
-                        "type": "image_url",
-                        "image_url": {
-                            "url": image_str,
-                        },
-                    },
-                ],
-            },
-        ],
-        "max_tokens": 300,  # Default max tokens
-    }
-
-    if token:
-        completion_kwargs["api_key"] = token
-    if base_url:
-        completion_kwargs["api_base"] = base_url
-    response = litellm.completion(**completion_kwargs, **kwargs)
-    return response.choices[0].message.content.strip()
-
-
 if __name__ == "__main__":
-    response = llm_analyze_image(
-        image_url="https://picsum.photos/200/300",
+    response = llm_generate_image(
+        prompt="A dog!",
         model="ollama/llava",
     )
     print(response)

src/jinjarope/resources/llm_filters.toml

@@ -1,9 +1,9 @@
 # This file contains definitions for functions of the "litellm" library.
 
 [filters.llm_complete]
-fn = "jinjarope.llmfilters.llm_complete"
+fn = "llmling_agent_functional.run_agent"
 group = "llm"
-required_packages = ["litellm"]
+required_packages = ["llmling-agent", "litellm"]
 
 [filters.llm_complete.examples.basic]
 template = """
@@ -23,24 +23,4 @@ template = """
 [filters.llm_generate_image]
 fn = "jinjarope.llmfilters.llm_generate_image"
 group = "llm"
-required_packages = ["litellm", "pillow"]
-
-# [filters.llm_generate_image.examples.basic]
-# template = """
-# {{ "A beautiful sunset over the ocean" | llm_generate_image(model="gemini/gemini-pro-vision") }}
-# """
-
-# [filters.llm_generate_image.examples.with_size]
-# template = """
-# {{ "A futuristic cityscape" | llm_generate_image(model="gemini/gemini-pro-vision", size="512x512") }}
-# """
-
-# [filters.llm_generate_image.examples.with_quality]
-# template = """
-# {{ "A detailed portrait of a cat" | llm_generate_image(model="gemini/gemini-pro-vision", quality="hd") }}
-# """
-
-# [filters.llm_generate_image.examples.as_b64_json]
-# template = """
-# {{ "A colorful abstract painting" | llm_generate_image(model="gemini/gemini-pro-vision", as_b64_json=true) }}
-# """
+required_packages = ["llmling-agent", "litellm", "pillow"]