pltr-cli 0.12.0__py3-none-any.whl → 0.13.0__py3-none-any.whl

pltr/commands/language_models.py

@@ -0,0 +1,515 @@
+"""
+LanguageModels commands for Foundry.
+Provides commands for Anthropic Claude models and OpenAI embeddings.
+"""
+
+import typer
+import json
+from typing import Optional, List, Any, TYPE_CHECKING
+from pathlib import Path
+from rich.console import Console
+
+# Lazy import to avoid SDK Literal types being processed by typer at module load time
+if TYPE_CHECKING:
+    pass
+
+from ..utils.formatting import OutputFormatter
+from ..utils.progress import SpinnerProgressTracker
+from ..auth.base import ProfileNotFoundError, MissingCredentialsError
+from ..utils.completion import (
+    complete_profile,
+    complete_output_format,
+)
+
+# Create main app and sub-apps
+app = typer.Typer(help="Interact with language models")
+anthropic_app = typer.Typer(help="Anthropic Claude models")
+openai_app = typer.Typer(help="OpenAI models")
+
+# Add sub-apps
+app.add_typer(anthropic_app, name="anthropic")
+app.add_typer(openai_app, name="openai")
+
+console = Console()
+formatter = OutputFormatter(console)
+
+
+def parse_json_input(input_str: str) -> Any:
+    """
+    Parse JSON input from string or file.
+
+    Supports:
+    - Inline JSON: '{"key": "value"}' or '["item1", "item2"]'
+    - File reference: @input.json
+
+    Args:
+        input_str: Input string or file reference
+
+    Returns:
+        Parsed JSON data
+
+    Raises:
+        FileNotFoundError: If file reference doesn't exist
+        json.JSONDecodeError: If JSON is invalid
+    """
+    if not input_str:
+        return None
+
+    # Handle file reference
+    if input_str.startswith("@"):
+        file_path = Path(input_str[1:])
+        if not file_path.exists():
+            raise FileNotFoundError(f"Input file not found: {file_path}")
+
+        with open(file_path, "r") as f:
+            return json.load(f)
+
+    # Handle inline JSON
+    return json.loads(input_str)
+
+
+def display_anthropic_response(response: dict, format: str, output: Optional[str]):
+    """
+    Display Anthropic model response with token usage.
+
+    Args:
+        response: Response dictionary from service
+        format: Output format
+        output: Optional output file path
+    """
+    # Display token usage prominently if available
+    if "usage" in response:
+        usage = response["usage"]
+        input_tokens = usage.get("inputTokens", 0)
+        output_tokens = usage.get("outputTokens", 0)
+        total_tokens = usage.get("totalTokens", input_tokens + output_tokens)
+
+        formatter.print_info(
+            f"Token usage - Input: {input_tokens}, Output: {output_tokens}, Total: {total_tokens}"
+        )
+
+    # Display full response
+    if output:
+        formatter.save_to_file(response, output, format)
+        formatter.print_success(f"Response saved to {output}")
+    else:
+        formatter.display(response, format)
+
+
+def display_openai_response(response: dict, format: str, output: Optional[str]):
+    """
+    Display OpenAI model response with token usage.
+
+    Args:
+        response: Response dictionary from service
+        format: Output format
+        output: Optional output file path
+    """
+    # Display token usage prominently if available
+    if "usage" in response:
+        usage = response["usage"]
+        prompt_tokens = usage.get("promptTokens", 0)
+        total_tokens = usage.get("totalTokens", prompt_tokens)
+
+        formatter.print_info(
+            f"Token usage - Prompt: {prompt_tokens}, Total: {total_tokens}"
+        )
+
+    # Display full response
+    if output:
+        formatter.save_to_file(response, output, format)
+        formatter.print_success(f"Response saved to {output}")
+    else:
+        formatter.display(response, format)
+
+
+# ===== Anthropic Commands =====
+
+
+@anthropic_app.command("messages")
+def anthropic_messages(
+    model_id: str = typer.Argument(
+        ...,
+        help="Model Resource Identifier (ri.language-models.main.model.<id>)",
+    ),
+    message: str = typer.Option(
+        ...,
+        "--message",
+        "-m",
+        help="User message text",
+    ),
+    max_tokens: int = typer.Option(
+        1024,
+        "--max-tokens",
+        help="Maximum tokens to generate",
+    ),
+    system: Optional[str] = typer.Option(
+        None,
+        "--system",
+        "-s",
+        help="System prompt to guide model behavior",
+    ),
+    temperature: Optional[float] = typer.Option(
+        None,
+        "--temperature",
+        "-t",
+        help="Sampling temperature (0.0-1.0). Lower is more deterministic.",
+        min=0.0,
+        max=1.0,
+    ),
+    stop: Optional[List[str]] = typer.Option(
+        None,
+        "--stop",
+        help="Stop sequences (can be specified multiple times)",
+    ),
+    top_k: Optional[int] = typer.Option(
+        None,
+        "--top-k",
+        help="Sample from top K tokens (Anthropic models only)",
+        min=1,
+    ),
+    top_p: Optional[float] = typer.Option(
+        None,
+        "--top-p",
+        help="Nucleus sampling threshold (0.0-1.0)",
+        min=0.0,
+        max=1.0,
+    ),
+    profile: Optional[str] = typer.Option(
+        None,
+        "--profile",
+        "-p",
+        help="Profile name",
+        autocompletion=complete_profile,
+    ),
+    format: str = typer.Option(
+        "json",
+        "--format",
+        "-f",
+        help="Output format (table, json, csv)",
+        autocompletion=complete_output_format,
+    ),
+    output: Optional[str] = typer.Option(
+        None, "--output", "-o", help="Output file path"
+    ),
+    preview: bool = typer.Option(
+        False,
+        "--preview",
+        help="Enable preview mode",
+    ),
+):
+    """
+    Send a single message to an Anthropic Claude model.
+
+    Simple interface for single-turn Q&A with Claude models. For multi-turn
+    conversations, tool calling, or advanced features, use messages-advanced.
+
+    Examples:
+
+        # Basic message
+        pltr language-models anthropic messages ri.language-models.main.model.abc123 \\
+            --message "Explain quantum computing"
+
+        # With system prompt and custom parameters
+        pltr language-models anthropic messages ri.language-models.main.model.abc123 \\
+            --message "Write a haiku" \\
+            --system "You are a poetic assistant" \\
+            --temperature 0.8 \\
+            --max-tokens 100
+
+        # With stop sequences
+        pltr language-models anthropic messages ri.language-models.main.model.abc123 \\
+            --message "List three items" \\
+            --stop "." --stop "\\n\\n"
+
+        # Save response to file
+        pltr language-models anthropic messages ri.language-models.main.model.abc123 \\
+            --message "Summarize AI trends" \\
+            --output response.json
+    """
+    try:
+        from ..services.language_models import LanguageModelsService
+
+        service = LanguageModelsService(profile=profile)
+
+        with SpinnerProgressTracker().track_spinner("Sending message..."):
+            response = service.send_message(
+                model_id=model_id,
+                message=message,
+                max_tokens=max_tokens,
+                system=system,
+                temperature=temperature,
+                stop_sequences=stop if stop else None,
+                top_k=top_k,
+                top_p=top_p,
+                preview=preview,
+            )
+
+        display_anthropic_response(response, format, output)
+
+    except (ProfileNotFoundError, MissingCredentialsError) as e:
+        formatter.print_error(f"Authentication error: {e}")
+        raise typer.Exit(1)
+    except (FileNotFoundError, json.JSONDecodeError) as e:
+        formatter.print_error(f"Invalid input: {e}")
+        raise typer.Exit(1)
+    except Exception as e:
+        formatter.print_error(f"Operation failed: {e}")
+        raise typer.Exit(1)
+
+
+@anthropic_app.command("messages-advanced")
+def anthropic_messages_advanced(
+    model_id: str = typer.Argument(
+        ...,
+        help="Model Resource Identifier (ri.language-models.main.model.<id>)",
+    ),
+    request: str = typer.Option(
+        ...,
+        "--request",
+        "-r",
+        help="Request JSON (inline or @file.json). Must include 'messages' and 'maxTokens'.",
+    ),
+    profile: Optional[str] = typer.Option(
+        None,
+        "--profile",
+        "-p",
+        help="Profile name",
+        autocompletion=complete_profile,
+    ),
+    format: str = typer.Option(
+        "json",
+        "--format",
+        "-f",
+        help="Output format (table, json, csv)",
+        autocompletion=complete_output_format,
+    ),
+    output: Optional[str] = typer.Option(
+        None, "--output", "-o", help="Output file path"
+    ),
+    preview: bool = typer.Option(
+        False,
+        "--preview",
+        help="Enable preview mode",
+    ),
+):
+    """
+    Send messages to Anthropic Claude model with advanced features.
+
+    Accepts full SDK request structure, enabling:
+    - Multi-turn conversations
+    - Tool/function calling
+    - Extended thinking mode
+    - Document and image processing
+    - Citations
+
+    The request must be a JSON object containing:
+    - messages: List of message objects with role and content
+    - maxTokens: Maximum tokens to generate
+
+    Optional fields:
+    - system: System prompt blocks
+    - temperature: Sampling temperature (0.0-1.0)
+    - thinking: Extended thinking configuration
+    - tools: Tool definitions for function calling
+    - toolChoice: Tool selection strategy
+    - stopSequences: Sequences that stop generation
+    - topK: Sample from top K tokens
+    - topP: Nucleus sampling threshold
+
+    Examples:
+
+        # Multi-turn conversation from file
+        # conversation.json:
+        # {
+        #   "messages": [
+        #     {"role": "user", "content": [{"type": "text", "text": "Hi"}]},
+        #     {"role": "assistant", "content": [{"type": "text", "text": "Hello!"}]},
+        #     {"role": "user", "content": [{"type": "text", "text": "Help me"}]}
+        #   ],
+        #   "maxTokens": 500
+        # }
+        pltr language-models anthropic messages-advanced ri.language-models.main.model.abc123 \\
+            --request @conversation.json
+
+        # Inline JSON with system prompt
+        pltr language-models anthropic messages-advanced ri.language-models.main.model.abc123 \\
+            --request '{"messages": [{"role": "user", "content": [{"type": "text", "text": "Hi"}]}], "maxTokens": 100, "system": [{"type": "text", "text": "Be concise"}]}'
+
+        # With extended thinking
+        pltr language-models anthropic messages-advanced ri.language-models.main.model.abc123 \\
+            --request '{"messages": [{"role": "user", "content": [{"type": "text", "text": "Solve this problem"}]}], "maxTokens": 2000, "thinking": {"type": "enabled", "budget": 10000}}'
+    """
+    try:
+        # Parse request JSON
+        request_data = parse_json_input(request)
+
+        # Validate required fields
+        if not isinstance(request_data, dict):
+            raise ValueError("Request must be a JSON object")
+        if "messages" not in request_data:
+            raise ValueError("Request must include 'messages' field")
+        if "maxTokens" not in request_data:
+            raise ValueError("Request must include 'maxTokens' field")
+
+        from ..services.language_models import LanguageModelsService
+
+        service = LanguageModelsService(profile=profile)
+
+        with SpinnerProgressTracker().track_spinner("Sending messages..."):
+            response = service.send_messages_advanced(
+                model_id=model_id,
+                messages=request_data["messages"],
+                max_tokens=request_data["maxTokens"],
+                system=request_data.get("system"),
+                temperature=request_data.get("temperature"),
+                thinking=request_data.get("thinking"),
+                tools=request_data.get("tools"),
+                tool_choice=request_data.get("toolChoice"),
+                stop_sequences=request_data.get("stopSequences"),
+                top_k=request_data.get("topK"),
+                top_p=request_data.get("topP"),
+                preview=preview,
+            )
+
+        display_anthropic_response(response, format, output)
+
+    except (ProfileNotFoundError, MissingCredentialsError) as e:
+        formatter.print_error(f"Authentication error: {e}")
+        raise typer.Exit(1)
+    except (FileNotFoundError, json.JSONDecodeError, ValueError) as e:
+        formatter.print_error(f"Invalid input: {e}")
+        raise typer.Exit(1)
+    except Exception as e:
+        formatter.print_error(f"Operation failed: {e}")
+        raise typer.Exit(1)
+
+
+# ===== OpenAI Commands =====
+
+
+@openai_app.command("embeddings")
+def openai_embeddings(
+    model_id: str = typer.Argument(
+        ...,
+        help="Model Resource Identifier (ri.language-models.main.model.<id>)",
+    ),
+    input: str = typer.Option(
+        ...,
+        "--input",
+        "-i",
+        help='Input text(s). Single string, JSON array \'["text1", "text2"]\', or @file.json',
+    ),
+    dimensions: Optional[int] = typer.Option(
+        None,
+        "--dimensions",
+        "-d",
+        help="Custom embedding dimensions (not all models support this)",
+        min=1,
+    ),
+    encoding: Optional[str] = typer.Option(
+        None,
+        "--encoding",
+        "-e",
+        help="Output encoding format: 'float' or 'base64'",
+    ),
+    profile: Optional[str] = typer.Option(
+        None,
+        "--profile",
+        "-p",
+        help="Profile name",
+        autocompletion=complete_profile,
+    ),
+    format: str = typer.Option(
+        "json",
+        "--format",
+        "-f",
+        help="Output format (table, json, csv)",
+        autocompletion=complete_output_format,
+    ),
+    output: Optional[str] = typer.Option(
+        None, "--output", "-o", help="Output file path"
+    ),
+    preview: bool = typer.Option(
+        False,
+        "--preview",
+        help="Enable preview mode",
+    ),
+):
+    """
+    Generate embeddings for text using an OpenAI model.
+
+    Accepts single text strings or multiple texts for batch processing.
+    Returns embedding vectors that can be used for semantic search,
+    clustering, or similarity comparisons.
+
+    Examples:
+
+        # Single text
+        pltr language-models openai embeddings ri.language-models.main.model.xyz789 \\
+            --input "Machine learning is fascinating"
+
+        # Multiple texts (inline JSON array)
+        pltr language-models openai embeddings ri.language-models.main.model.xyz789 \\
+            --input '["Document 1", "Document 2", "Document 3"]'
+
+        # Multiple texts from file
+        # texts.json: ["Text 1", "Text 2", "Text 3"]
+        pltr language-models openai embeddings ri.language-models.main.model.xyz789 \\
+            --input @texts.json
+
+        # Custom dimensions and encoding
+        pltr language-models openai embeddings ri.language-models.main.model.xyz789 \\
+            --input "Sample text" \\
+            --dimensions 1024 \\
+            --encoding base64
+
+        # Save embeddings to file
+        pltr language-models openai embeddings ri.language-models.main.model.xyz789 \\
+            --input '["Text 1", "Text 2"]' \\
+            --output embeddings.json
+    """
+    try:
+        # Parse input - handle string or array
+        input_data = (
+            parse_json_input(input)
+            if input.startswith("@") or input.startswith("[")
+            else input
+        )
+
+        # Convert to list if single string
+        if isinstance(input_data, str):
+            input_texts = [input_data]
+        elif isinstance(input_data, list):
+            input_texts = input_data
+        else:
+            raise ValueError("Input must be a string or array of strings")
+
+        # Validate encoding format if provided
+        if encoding and encoding not in ["float", "base64"]:
+            raise ValueError("Encoding format must be 'float' or 'base64'")
+
+        from ..services.language_models import LanguageModelsService
+
+        service = LanguageModelsService(profile=profile)
+
+        with SpinnerProgressTracker().track_spinner("Generating embeddings..."):
+            response = service.generate_embeddings(
+                model_id=model_id,
+                input_texts=input_texts,
+                dimensions=dimensions,
+                encoding_format=encoding,
+                preview=preview,
+            )
+
+        display_openai_response(response, format, output)
+
+    except (ProfileNotFoundError, MissingCredentialsError) as e:
+        formatter.print_error(f"Authentication error: {e}")
+        raise typer.Exit(1)
+    except (FileNotFoundError, json.JSONDecodeError, ValueError) as e:
+        formatter.print_error(f"Invalid input: {e}")
+        raise typer.Exit(1)
+    except Exception as e:
+        formatter.print_error(f"Operation failed: {e}")
+        raise typer.Exit(1)