Perplexity 推出 Rust 检索引擎 Photon，p99 延迟从 800 ms 降至 65 ms
Perplexity Introduces Photon: A Rust-Based Retrieval Engine That Cuts p99 Latency From 800 ms to 65 ms - MarkTechPost
Perplexity 发布自研检索与排序引擎 Photon，用 Rust 编写，取代此前 fork 的开源引擎，现已承载全部生产流量，并支撑 Perplexity Search API 新增的 Fast Search 模式。官方称单次调用延迟为 p50 160 ms、p95 230 ms，p99 从 800 ms 降至 65 ms，完整网页索引可在个位数小时内构建完成。Fast Search 以 search_type: "fast" 通过托管 API 提供，每 1000 次请求收费 1 美元，Photon 本身不开源、无法自托管；在 6 个基准共 3554 个任务上得分 64.3%、成本约 59.73 美元，默认预设为 64.0%、187.60 美元，但长尾相关性 DCG 从 2.45 降至 2.21，答案可得率下降 2.9 个百分点。
推荐理由
AI 评分 67
Perplexity 自研检索引擎 Photon 的延迟与成本数据，以及 Fast Search 在质量上的取舍，可供做搜索或智能体检索的人参考。
主题
正文
当前显示原文，中文全文译文尚未就绪。
Perplexity has released Photon , an in-house retrieval and ranking engine written in Rust. It replaces an open-source engine Perplexity had forked for its AI-native search stack. Photon now handles retrieval and ranking for all production traffic. It also powers a new Fast Search mode in the Perplexity Search API. Perplexity reports single-call latency of 160 ms at p50 and 230 ms at p95.
Is it deployable? Yes, as a hosted API. Set search_type: "fast" on POST /search and pay $1 per 1,000 requests. Photon itself is not open source, so the engine cannot be self-hosted.
The old engine hit 3 limits as the index grew:
Perplexity team concluded that building from scratch was simpler and cheaper than maintaining its fork.
A load balancer routes each request to a Photon broker. The broker fans out to a shard group and watches for timeouts. Each shard runs retrieval, initial ranking, and second-stage ranking. The broker then merges candidates and fetches key document fields.
A full web index now builds in a single-digit number of hours.
Fast Search pairs Photon with lighter ranking tuned for agentic workflows. Perplexity tested it on 6 benchmarks: WideSearch, BrowseComp , DSQA, FRAMES, SEAL-0, and SEAL-Hard. Across 3,554 tasks, Fast scored 64.3% at $59.73 in estimated model-plus-search cost. The default preset scored 64.0% at $187.60, so Fast was about 68% cheaper.
The trade-off shows up in broader search quality. On internal long-tail benchmarks, relevance (DCG) fell from 2.45 to 2.21. Answer availability dropped from 0.596 to 0.567, a loss of 2.9 percentage points. Perplexity recommends Fast for day-to-day agent loops and the default for hard, ambiguous queries.
curl -X POST 'https://api.perplexity.ai/search' \ -H "Authorization: Bearer $PERPLEXITY_API_KEY" \ -H 'Content-Type: application/json' \ -d '{"query": "latest stable Rust release", "search_type": "fast", "max_results": 5}'
On Python SDK 0.43.4 and 0.43.5, pass extra_body={"search_type": "fast"} per the docs .
All latency figures are vendor-reported under different setups, so they are not like-for-like.