Files
openclaw/extensions/parallel/src/parallel-web-search-provider.ts
Peter Steinberger fa77fe10d5 chore: migrate active GPT-5.5 references to GPT-5.6 (#104452)
* chore(models): migrate active GPT-5.5 references

* test(workboard): expect GPT-5.6 Sol default

* chore: keep release notes in PR body

* test(models): align picker fixtures with Sol default

* test: update PDF default model expectation

* test(qa): migrate thinking smoke to Luna

* test(gateway): align mock catalog with Sol default

* ci: retrigger exact-head PR checks

* test(gateway): document default catalog invariant
2026-07-11 06:30:57 -07:00

72 lines
3.4 KiB
TypeScript

import { createLazyRuntimeModule } from "openclaw/plugin-sdk/lazy-runtime";
import type { WebSearchProviderPlugin } from "openclaw/plugin-sdk/provider-web-search-contract";
import { createParallelWebSearchProviderBase } from "./parallel-web-search-provider.shared.js";
const PARALLEL_MAX_SEARCH_COUNT = 40;
const PARALLEL_MAX_SEARCH_QUERIES = 5;
const PARALLEL_MAX_SEARCH_QUERY_CHARS = 200;
const PARALLEL_MAX_OBJECTIVE_CHARS = 5000;
const PARALLEL_MAX_SESSION_ID_CHARS = 1000;
const PARALLEL_MAX_CLIENT_MODEL_CHARS = 100;
const loadParallelWebSearchRuntime = createLazyRuntimeModule(
() => import("./parallel-web-search-provider.runtime.js"),
);
// Mirrors Parallel's recommended search tool schema:
// https://docs.parallel.ai/search/best-practices#search-tool-definition
export const ParallelSearchSchema = {
type: "object",
properties: {
objective: {
type: "string",
description:
"Natural-language description of the underlying question or goal driving the search. Should be self-contained with enough context to understand the intent. Used together with search_queries to focus results on the most relevant content.",
maxLength: PARALLEL_MAX_OBJECTIVE_CHARS,
},
search_queries: {
type: "array",
description:
"Concise keyword search queries, 3-6 words each. Provide 2-3 diverse queries for best results (max 5). Vary entity names, synonyms, and angles. Each query is a keyword phrase, not a sentence; do not use site: operators.",
items: { type: "string", maxLength: PARALLEL_MAX_SEARCH_QUERY_CHARS },
minItems: 1,
maxItems: PARALLEL_MAX_SEARCH_QUERIES,
},
count: {
type: "integer",
description: "Number of results to return (1-40).",
minimum: 1,
maximum: PARALLEL_MAX_SEARCH_COUNT,
},
session_id: {
type: "string",
description:
"Optional session id returned by an earlier Parallel search. Pass it on follow-up searches that are part of the same task to keep Parallel's server-side context grouped (look for `sessionId` in the prior tool result).",
maxLength: PARALLEL_MAX_SESSION_ID_CHARS,
},
client_model: {
type: "string",
description:
"The identifier of the LLM model making this tool call (e.g. 'claude-opus-4-7', 'gpt-5.6-sol', 'gemini-3.1-pro'). Pass the exact active model slug verbatim; never shorten or substitute a family alias like 'gpt-5'. Lets Parallel tailor default settings for your model's capabilities.",
maxLength: PARALLEL_MAX_CLIENT_MODEL_CHARS,
},
},
required: ["objective", "search_queries"],
additionalProperties: false,
} satisfies Record<string, unknown>;
export function createParallelWebSearchProvider(): WebSearchProviderPlugin {
return {
...createParallelWebSearchProviderBase(),
createTool: (ctx) => ({
description:
"Search the web using Parallel. Returns ranked, LLM-optimized dense excerpts from web sources. Pass an `objective` describing the underlying question along with 2-3 short keyword `search_queries` (Parallel's recommended pairing). For multi-step research, thread the prior result's `sessionId` back in as `session_id` to keep Parallel's context grouped.",
parameters: ParallelSearchSchema,
execute: async (args) => {
const { executeParallelWebSearchProviderTool } = await loadParallelWebSearchRuntime();
return await executeParallelWebSearchProviderTool(ctx, args);
},
}),
};
}