Define refinement actions in the model providers (#87)

* Add descriptions of fields in llama 2 input schema

* Let GPT-4 know when the provider stays the same

* Allow refetching in the event of any errors

* Define refinement actions in model providers

* Fix prettier

src/modelProviders/replicate-llama2/index.ts

@@ -38,26 +38,42 @@ const modelProvider: ReplicateLlama2Provider = {
         type: "string",
         enum: supportedModels as unknown as string[],
       },
+      system_prompt: {
+        type: "string",
+        description:
+          "System prompt to send to Llama v2. This is prepended to the prompt and helps guide system behavior.",
+      },
       prompt: {
         type: "string",
+        description: "Prompt to send to Llama v2.",
       },
       stream: {
         type: "boolean",
+        description: "Whether to stream output from Llama v2.",
       },
-      max_length: {
+      max_new_tokens: {
         type: "number",
+        description:
+          "Maximum number of tokens to generate. A word is generally 2-3 tokens (minimum: 1)",
       },
       temperature: {
         type: "number",
+        description:
+          "Adjusts randomness of outputs, greater than 1 is random and 0 is deterministic, 0.75 is a good starting value. (minimum: 0.01; maximum: 5)",
       },
       top_p: {
         type: "number",
+        description:
+          "When decoding text, samples from the top p percentage of most likely tokens; lower to ignore less likely tokens (minimum: 0.01; maximum: 1)",
       },
       repetition_penalty: {
         type: "number",
+        description:
+          "Penalty for repeated words in generated text; 1 is no penalty, values greater than 1 discourage repetition, less than 1 encourage it. (minimum: 0.01; maximum: 5)",
       },
       debug: {
         type: "boolean",
+        description: "provide debugging output in logs",
       },
     },
     required: ["model", "prompt"],