Perplexity - chat-completions-advanced

components/perplexity/actions/chat-completions-advanced/chat-completions-advanced.mjs

@@ -0,0 +1,185 @@
+import app from "../../perplexity.app.mjs";
+
+export default {
+  key: "perplexity-chat-completions-advanced",
+  name: "Chat Completions (Advanced)",
+  description: "Generates a model's response for the given chat conversation with multi-message support and Perplexity search controls. Docs: https://docs.perplexity.ai/api-reference/chat-completions-post",
+  version: "0.0.1",
+  annotations: {
+    destructiveHint: false,
+    openWorldHint: true,
+    readOnlyHint: false,
+  },
+  type: "action",
+  props: {
+    app,
+    model: {
+      propDefinition: [
+        app,
+        "model",
+      ],
+    },
+    messages: {
+      type: "string[]",
+      label: "Messages (JSON strings or UI collection)",
+      description: "Array of message objects: [{ role: 'system'|'user'|'assistant', content: '...' }, ...]. If provided, 'role' and 'content' are ignored.",
+      optional: true,
+    },
+    system: {
+      type: "string",
+      label: "System",
+      description: "Optional system prompt injected as the first message",
+      optional: true,
+    },
+    role: {
+      propDefinition: [
+        app,
+        "role",
+      ],
+      optional: true,
+    },
+    content: {
+      propDefinition: [
+        app,
+        "content",
+      ],
+      optional: true,
+    },
+    temperature: {
+      type: "string",
+      label: "Temperature",
+      description: "Sampling temperature. Higher values = more diverse output.",
+      optional: true,
+    },
+    topP: {
+      type: "string",
+      label: "Top-p",
+      description: "Nucleus sampling probability mass. Use either temperature or top_p.",
+      optional: true,
+    },
+    maxTokens: {
+      type: "integer",
+      label: "Max Output Tokens",
+      description: "Cap response length (Sonar Pro practical max output ~8k).",
+      optional: true,
+    },
+    stream: {
+      type: "boolean",
+      label: "Stream",
+      description: "Enable server-side streaming. This action will still buffer and return the final text.",
+      optional: true,
+      default: false,
+    },
+    searchDomainFilter: {
+      type: "string[]",
+      label: "Search Domain Filter",
+      description: "Limit web search to these domains (e.g., ['sec.gov','ft.com'])",
+      optional: true,
+    },
+    searchRecencyFilter: {
+      type: "string",
+      label: "Search Recency Filter",
+      description: "Prefer recent sources (e.g., 'day', 'week', 'month', 'year')",
+      optional: true,
+    },
+    topK: {
+      type: "integer",
+      label: "Top K",
+      description: "Restrict number of retrieved items considered",
+      optional: true,
+    },
+    returnImages: {
+      type: "boolean",
+      label: "Return Images",
+      description: "Ask API to include images when applicable",
+      optional: true,
+      default: false,
+    },
+    returnRelatedQuestions: {
+      type: "boolean",
+      label: "Return Related Questions",
+      description: "Ask API to include related questions in response",
+      optional: true,
+      default: false,
+    },
+  },
+
+  async run({ $ }) {
+    // Build messages array
+    let messages = [];
+    const provided = this.messages && this.messages.length > 0;
+
+    if (provided) {
+      // Accept messages as array of either objects or JSON strings
+      messages = this.messages.map((m) =>
+        typeof m === "string"
+          ? JSON.parse(m)
+          : m);
+    } else {
+      if (!this.content) {
+        throw new Error("Either provide `messages` or `content`.");
+      }
+      // Back-compat single-turn
+      if (this.system) {
+        messages.push({
+          role: "system",
+          content: this.system,
+        });
+      }
+      messages.push({
+        role: this.role || "user",
+        content: this.content,
+      });
+    }
+
+    const data = {
+      model: this.model,
+      messages,
+      // Generation knobs
+      ...(this.temperature != null && {
+        temperature: +this.temperature,
+      }),
+      ...(this.topP != null && {
+        top_p: +this.topP,
+      }),
+      ...(this.maxTokens != null && {
+        max_tokens: this.maxTokens,
+      }),
+      ...(this.stream != null && {
+        stream: this.stream,
+      }),
+      // Perplexity search controls
+      ...(this.searchDomainFilter && {
+        search_domain_filter: this.searchDomainFilter,
+      }),
+      ...(this.searchRecencyFilter && {
+        search_recency_filter: this.searchRecencyFilter,
+      }),
+      ...(this.topK != null && {
+        top_k: this.topK,
+      }),
+      ...(this.returnImages != null && {
+        return_images: this.returnImages,
+      }),
+      ...(this.returnRelatedQuestions != null && {
+        return_related_questions: this.returnRelatedQuestions,
+      }),
+    };
+
+    const response = await this.app.chatCompletions({
+      $,
+      data,
+    });
+
+    $.export(
+      "$summary",
+      `Perplexity ${this.model} responded${
+        this.stream
+          ? " (streaming buffered)"
+          : ""
+      }`,
+    );
+
+    return response;
+  },
+};

components/perplexity/actions/chat-completions/chat-completions.mjs

@@ -3,8 +3,8 @@ import app from "../../perplexity.app.mjs";
 export default {
   key: "perplexity-chat-completions",
   name: "Chat Completions",
-  description: "Generates a model's response for the given chat conversation. [See the documentation](https://docs.perplexity.ai/reference/post_chat_completions)",
-  version: "0.0.6",
+  description: "Generates a model's response for the given chat conversation. [See the documentation](https://docs.perplexity.ai/api-reference/chat-completions-post)",
+  version: "0.0.7",
   annotations: {
     destructiveHint: false,
     openWorldHint: true,

components/perplexity/package.json

@@ -13,6 +13,6 @@
     "access": "public"
   },
   "dependencies": {
-    "@pipedream/platform": "^1.6.5"
+    "@pipedream/platform": "^3.1.0"
   }
 }