Anthropic LLM¶
Runs prompts against Anthropic models via the LiteLLM integration. You provide a model, system prompt, user prompt, and generation settings; the node returns the generated text. Includes a built-in fallback mapping for common Anthropic model names.
Usage¶
Use this node when you need text generation, reasoning, or assistant-style responses from Anthropic’s Claude family. Typical workflow: choose a Claude model, set a system prompt to define behavior, pass your prompt content (optionally composed from other node outputs via input_1–input_4), tune temperature and max_tokens, then feed the resulting text to downstream nodes for parsing or display.
Inputs¶
| Field | Required | Type | Description | Example |
|---|---|---|---|---|
| model | True | STRING | Anthropic model to use. If the service cannot resolve the exact name, the node uses an internal fallback mapping for well-known Claude models. | claude-3-7-sonnet-20250219 |
| system_prompt | True | STRING | System instructions that steer the model’s role, tone, and constraints. | You are a helpful assistant. Be concise and cite sources when possible. |
| prompt | True | STRING | Main user prompt content to send to the model. You can compose this with optional inputs. | Summarize the following report and list three key risks. |
| temperature | True | FLOAT | Controls randomness. Lower is more deterministic; higher is more creative. | 0.5 |
| max_tokens | True | INT | Maximum number of tokens to generate in the response. | 1024 |
| input_1 | False | STRING | Optional extra text input for composing prompts (e.g., context, retrieved passages). | Previous conversation transcript |
| input_2 | False | STRING | Optional extra text input for composing prompts. | User profile or preferences |
| input_3 | False | STRING | Optional extra text input for composing prompts. | Knowledge base excerpt |
| input_4 | False | STRING | Optional extra text input for composing prompts. | Structured context or instructions |
Outputs¶
| Field | Type | Description | Example |
|---|---|---|---|
| Output | STRING | The model’s generated text output. | Here’s a concise summary of the report along with three key risks... |
Important Notes¶
- Model options include common Anthropic Claude variants and are backed by an internal fallback map. If a model alias is used, it resolves to a provider-qualified name like anthropic/claude-... depending on availability.
- Token limits and supported parameters can vary by model. Ensure max_tokens fits within the model’s context window together with your input size.
- Temperature meaning: lower values yield more focused, repeatable outputs; higher values increase creativity and variability.
- Optional inputs (input_1–input_4) are not merged automatically; construct your final prompt string accordingly in the prompt field if you need to include them.
- Access to Anthropic requires proper service configuration and credentials managed by the platform; ensure they are set by your administrator.
- This node returns a single STRING response; if you need structured output, instruct the model to format JSON and parse it in subsequent nodes.
Troubleshooting¶
- Invalid model name or unsupported version: Pick a model from the available options or use one of the mapped Claude names (e.g., claude-3-7-sonnet-20250219).
- Authorization or key errors: Verify that Anthropic access is configured by your administrator. The node itself does not accept keys directly.
- Context length or token limit exceeded: Reduce prompt size, lower max_tokens, or choose a model with a larger context window.
- Empty or low-quality outputs: Decrease temperature for more deterministic results, refine the system_prompt, and provide clearer instructions.
- Rate limits or timeouts: Retry with backoff, reduce request frequency, or simplify prompts to lower latency.