SKILL: Anthropic Provider Agent (Claude Backend)

Overview
The Provider Mandate
Symbolic Implementation (The Logic)
- Architectural Intent: Backend Request Execution
- Architectural Intent: Model Discovery
Registration

Overview

The Anthropic Provider Agent integrates Anthropic's Claude family of models as a pluggable System 1 (neural) backend for the Neurosymbolic Kernel. It enables the agent to utilize high-intelligence models for complex reasoning, drafting, and analysis tasks.

The Provider Mandate

Connectivity: Ensure reliable communication with the Anthropic Messages API.
Configurability: Dynamically resolve the specific Claude model to use based on the system's Environment Configuration.
Context Management: Leverage Claude's large context windows (up to 200k) for deep codebase analysis.
Safety: Handle API failures and missing credentials gracefully by logging errors rather than crashing the kernel.

Symbolic Implementation (The Logic)

The implementation focuses on the API request/response cycle and model discovery.

Architectural Intent: Backend Request Execution

This function handles the physical I/O with the Anthropic API. It retrieves credentials from the environment, model settings from the Config skill, and formats the request according to the Anthropic Messages protocol.

(defun execute-anthropic-request (prompt system-prompt)
  "Executes a completion request via the Anthropic (Claude) API."
  (let ((api-key (org-agent::get-env "ANTHROPIC_API_KEY"))
        (config-pkg (find-package :org-agent.skills.skill-environment-config)))
    (unless api-key
      (return-from execute-anthropic-request "(:type :LOG :payload (:text \"Anthropic key missing\"))"))
    
    (let* ((get-config-fn (when config-pkg (find-symbol "GET-CONFIG-ATTRIBUTE" config-pkg)))
           (model (if (and get-config-fn (fboundp get-config-fn))
                      (funcall get-config-fn :LLM_MODEL_ANTHROPIC "claude-3-5-sonnet-20240620")
                      "claude-3-5-sonnet-20240620"))
           (url "https://api.anthropic.com/v1/messages")
           (body (cl-json:encode-json-to-string
                  `((model . ,model)
                    (max_tokens . 1024)
                    (system . ,system-prompt)
                    (messages . (((role . "user") (content . ,prompt))))))))
      (handler-case
          (let* ((response (dex:post url
                                     :headers `(("Content-Type" . "application/json")
                                                ("x-api-key" . ,api-key)
                                                ("anthropic-version" . "2023-06-01"))
                                     :content body))
                 (json (cl-json:decode-json-from-string response)))
            ;; Extract content from Anthropic response
            (cdr (assoc :text (car (cdr (assoc :content json))))))
        (error (c)
          (format nil "(:type :LOG :payload (:text \"Anthropic Failure (~a) - ~a\"))" model c))))))

;; Register the backend
(org-agent:register-neuro-backend :anthropic #'execute-anthropic-request)
(org-agent:register-neuro-backend :claude #'execute-anthropic-request)

Architectural Intent: Model Discovery

Provides the Model Explorer with a list of supported models and their context limits, allowing the user to inspect capabilities.

(defun get-available-models ()
  "Returns the list of LLM models supported by this provider."
  '((:id "claude-3-5-sonnet-20240620" :context "200k")
    (:id "claude-3-opus-20240229" :context "200k")
    (:id "claude-3-haiku-20240307" :context "200k")))

Registration

(defskill :skill-provider-anthropic
  :priority 100
  :trigger (lambda (context) nil)
  :neuro (lambda (context) nil)
  :symbolic (lambda (action context) action))

3.7 KiB Raw Blame History