@apicity/ 消防工作
Fireworks AI供应商用于聊天完成,完成,并嵌入.
运行时间依赖性 :
zod@^4.6.6- 请求在每一个 POST 端点上附加计划.schema
安装
npm install @apicity/fireworks
# or
pnpm add @apicity/fireworks
快速启动
import { createFireworks } from "@apicity/fireworks";
const fireworks = createFireworks({ apiKey: process.env.FIREWORKS_API_KEY! });
真实世界的例子: 重新排序候选段落, 然后流出一个 Llama 3.3 70B 答案
烟花飞来两道推论表面 清净地组成一个
轻量级RAG管道:主机跨编码器重排(qwen3-reranker-8b(中文(简体) )
对 N 候选文件的查询进行评分,并且 OpenAI 兼容
llama 3.3 70B Instruct上的流式聊天完成. 重排是
硬部分——大多数提供者只销售嵌入式——和流式聊天
是个便宜又快的尾巴 镜下两步流
tests/integration/fireworks-rerank.test.ts
财务报告和审定财务报表
tests/integration/fireworks-stream.test.ts, (中文(简体) ),
反弹
tests/recordings/fireworks_626462085/rerank-basic_678261817/
财务报告和审定财务报表
tests/recordings/fireworks_626462085/chat-stream-hello_2801342443/, (中文(简体) ),
因此,每个分数,符号计数,和下面的SSE形状直接来自
记录的HARs.
import { createFireworks } from "@apicity/fireworks";
import type {
FireworksRerankResponse,
FireworksChatStreamChunk,
} from "@apicity/fireworks";
const fireworks = createFireworks({ apiKey: process.env.FIREWORKS_API_KEY! });
// 1. Cross-encoder rerank. The model sees the query and each document
// jointly (unlike embedding-based retrieval, which scores them
// independently), so it picks up phrasing-level signal that ANN
// search misses. `top_n` caps the response length; `return_documents:
// true` echoes the document text back inline so you can hand the
// top hit straight to a chat model without keeping a parallel
// id→text map.
const ranked: FireworksRerankResponse =
await fireworks.inference.v1.rerank({
model: "fireworks/qwen3-reranker-8b",
query: "What is the capital of France?",
documents: [
"Berlin is the capital of Germany.",
"Paris is the capital and largest city of France.",
"Madrid is the capital of Spain.",
],
top_n: 2,
return_documents: true,
});
// 2. The response is sorted descending by relevance_score. The score
// spread is huge — Paris ~0.96, Berlin ~0.0001, Madrid dropped — a
// near-four-orders-of-magnitude gap that a cosine-similarity
// retriever could never produce. `index` points back into the
// original `documents` array.
const top = ranked.data[0];
console.log(`#1: index=${top.index}, score=${top.relevance_score.toFixed(4)}`);
console.log(` "${top.document}"`);
// → "#1: index=1, score=0.9579"
// → "Paris is the capital and largest city of France."
console.log(`model=${ranked.model}, tokens=${ranked.usage.prompt_tokens}`);
// → "model=accounts/fireworks/models/qwen3-reranker-8b, tokens=261"
// 3. Stream a chat answer through Llama 3.3 70B Instruct. The recorded
// call below is a generic 'say hello' smoke test — independent from
// the rerank above — but the request shape is exactly what you'd
// send in production: drop `top.document` into a system message and
// the user's original question into the user turn.
// `temperature: 0` makes the output deterministic; `max_tokens: 64`
// caps the response. The streaming surface lives under
// `post.stream.*` and yields OpenAI-shaped
// `chat.completion.chunk` objects.
let answer = "";
let finish: string | null = null;
let totals = { prompt: 0, completion: 0 };
const stream = fireworks.post.stream.inference.v1.chat.completions({
model: "accounts/fireworks/models/llama-v3p3-70b-instruct",
messages: [{ role: "user", content: "Say hello in one sentence." }],
temperature: 0,
max_tokens: 64,
});
for await (const chunk of stream) {
// 4. Each delta carries one or two tokens of `content`. The FIRST
// delta carries `delta.role` and no content; the LAST delta has
// no content but `finish_reason: "stop"` and final `usage`.
// Read defensively — both fields are optional on every chunk.
const choice: FireworksChatStreamChunk["choices"][number] | undefined =
chunk.choices[0];
if (choice?.delta?.content) answer += choice.delta.content;
if (choice?.finish_reason) finish = choice.finish_reason;
if (chunk.usage) {
totals = {
prompt: chunk.usage.prompt_tokens,
completion: chunk.usage.completion_tokens,
};
}
}
console.log(answer);
// → "Hello, it's nice to meet you and I'm here to help with any questions or topics you'd like to discuss."
console.log(
`finish=${finish}, prompt=${totals.prompt}, completion=${totals.completion}`,
);
// → "finish=stop, prompt=41, completion=26"
页:1
- 重排是交叉编码器,不是嵌入模型——它
直接计算查询-文档交互,所以它实质上是
比同位素相似性更准确 而不是独立的嵌入,
但它并没有扩大到数百万候选人。 标准混合
是基于嵌入的 ANN 来获取一个上- 100, 那么
qwen3-reranker-8b精炼到顶端3。 烟花主机都一半在同一地 域( 选项)/v1/embeddings第一阶段,/v1/rerank联 合 国 页:1. - 流出聊天块是 OpenAI 兼容的 :
chat.completion.chunk与choices[].delta含有增量内容。 看着finish_reason想知道你是否中弹"stop", (中文(简体) ),"length",或"content_filter"只有 最终 块有usage人口; 中间块有usage: null烟花也露出a 每一次通话fireworks-cached-prompt-tokens回应头-40 上面的41个缓存, 前缀几乎全部支付, 可见而不分析身体. - 对同一Llama目录的Anthropic形状的反应,互换
改为
fireworks.inference.v1.messages({...})模式标识和定价 匹配,但反应是{ role, content: [{ type: "text", text }], ... }改为{ choices: [{ message: { content }}] }- 当 移植已写入的代码@apicity/anthropic。 。 。 . - 对于非分流完成,请放下
post.stream.前缀 :fireworks.inference.v1.chat.completions({...})返回单个FireworksChatResponse与choices[0].message.content人口数 完整usage在一枪。 相同的有效载荷计划,在fireworks.inference.v1.chat.completions.schema运行时间 在呼叫前验证. - 丢出错误
FireworksError与status和被剖析出的身体 页:1.400从重新排序通常表示空/ 丢失documents阵列;429是费率限制;503是部署冷 开始 与withRetry({ retries: 3 })从@apicity/fireworks没有密码的冷部署.
API 参考
98个终点跨一组. 每个方法都映射出上游的 URL 路径 .
推论
POST fireworks.inference.v1.accounts.apiKeys.create
POST https://api.fireworks.ai/v1/accounts/{accountId}/users/{userId}/apiKeys
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.apiKeys.create({ /* ... */ });
DELETE fireworks.inference.v1.accounts.apiKeys
DELETE https://api.fireworks.ai/v1/accounts/{accountId}/users/{userId}/apiKeys:delete
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.apiKeys.delete({ /* ... */ });
GET fireworks.inference.v1.accounts.apiKeys.list
GET https://api.fireworks.ai/v1/accounts/{accountId}/users/{userId}/apiKeys
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.apiKeys.list({ /* ... */ });
POST fireworks.inference.v1.accounts.batchInferenceJobs.create
POST https://api.fireworks.ai/v1/accounts/{accountId}/batchInferenceJobs
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.batchInferenceJobs.create({ /* ... */ });
DELETE fireworks.inference.v1.accounts.batchInferenceJobs
DELETE https://api.fireworks.ai/v1/accounts/{accountId}/batchInferenceJobs/{jobId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.batchInferenceJobs.delete({ /* ... */ });
GET fireworks.inference.v1.accounts.batchInferenceJobs
GET https://api.fireworks.ai/v1/accounts/{accountId}/batchInferenceJobs/{jobId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.batchInferenceJobs.get({ /* ... */ });
GET fireworks.inference.v1.accounts.batchInferenceJobs.list
GET https://api.fireworks.ai/v1/accounts/{accountId}/batchInferenceJobs
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.batchInferenceJobs.list({ /* ... */ });
POST fireworks.inference.v1.accounts.datasets.create
POST https://api.fireworks.ai/v1/accounts/{accountId}/datasets
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.datasets.create({ /* ... */ });
DELETE fireworks.inference.v1.accounts.datasets
DELETE https://api.fireworks.ai/v1/accounts/{accountId}/datasets/{datasetId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.datasets.delete({ /* ... */ });
GET fireworks.inference.v1.accounts.datasets
GET https://api.fireworks.ai/v1/accounts/{accountId}/datasets/{datasetId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.datasets.get({ /* ... */ });
GET fireworks.inference.v1.accounts.datasets.getDownloadEndpoint
GET https://api.fireworks.ai/v1/accounts/{accountId}/datasets/{datasetId}:getDownloadEndpoint
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.datasets.getDownloadEndpoint({ /* ... */ });
POST fireworks.inference.v1.accounts.datasets.getUploadEndpoint
POST https://api.fireworks.ai/v1/accounts/{accountId}/datasets/{datasetId}:getUploadEndpoint
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.datasets.getUploadEndpoint({ /* ... */ });
GET fireworks.inference.v1.accounts.datasets.list
GET https://api.fireworks.ai/v1/accounts/{accountId}/datasets
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.datasets.list({ /* ... */ });
PATCH fireworks.inference.v1.accounts.datasets.update
PATCH https://api.fireworks.ai/v1/accounts/{accountId}/datasets/{datasetId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.datasets.update({ /* ... */ });
POST fireworks.inference.v1.accounts.datasets.validateUpload
POST https://api.fireworks.ai/v1/accounts/{accountId}/datasets/{datasetId}:validateUpload
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.datasets.validateUpload({ /* ... */ });
POST fireworks.inference.v1.accounts.deployedModels.create
POST https://api.fireworks.ai/v1/accounts/{accountId}/deployedModels
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.deployedModels.create({ /* ... */ });
DELETE fireworks.inference.v1.accounts.deployedModels
DELETE https://api.fireworks.ai/v1/accounts/{accountId}/deployedModels/{deployedModelId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.deployedModels.delete({ /* ... */ });
GET fireworks.inference.v1.accounts.deployedModels
GET https://api.fireworks.ai/v1/accounts/{accountId}/deployedModels/{deployedModelId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.deployedModels.get({ /* ... */ });
GET fireworks.inference.v1.accounts.deployedModels.list
GET https://api.fireworks.ai/v1/accounts/{accountId}/deployedModels
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.deployedModels.list({ /* ... */ });
PATCH fireworks.inference.v1.accounts.deployedModels.update
PATCH https://api.fireworks.ai/v1/accounts/{accountId}/deployedModels/{deployedModelId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.deployedModels.update({ /* ... */ });
GET fireworks.inference.v1.accounts.deploymentShapes
GET https://api.fireworks.ai/v1/accounts/{accountId}/deploymentShapes/{shapeId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.deploymentShapes.get({ /* ... */ });
GET fireworks.inference.v1.accounts.deploymentShapes.versions
GET https://api.fireworks.ai/v1/accounts/{accountId}/deploymentShapes/{shapeId}/versions/{versionId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.deploymentShapes.versions.get({ /* ... */ });
GET fireworks.inference.v1.accounts.deploymentShapes.versions.list
GET https://api.fireworks.ai/v1/accounts/{accountId}/deploymentShapes/{shapeId}/versions
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.deploymentShapes.versions.list({ /* ... */ });
POST fireworks.inference.v1.accounts.deployments.create
POST https://api.fireworks.ai/v1/accounts/{accountId}/deployments
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.deployments.create({ /* ... */ });
DELETE fireworks.inference.v1.accounts.deployments
DELETE https://api.fireworks.ai/v1/accounts/{accountId}/deployments/{deploymentId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.deployments.delete({ /* ... */ });
GET fireworks.inference.v1.accounts.deployments
GET https://api.fireworks.ai/v1/accounts/{accountId}/deployments/{deploymentId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.deployments.get({ /* ... */ });
GET fireworks.inference.v1.accounts.deployments.list
GET https://api.fireworks.ai/v1/accounts/{accountId}/deployments
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.deployments.list({ /* ... */ });
PATCH fireworks.inference.v1.accounts.deployments.scale
PATCH https://api.fireworks.ai/v1/accounts/{accountId}/deployments/{deploymentId}:scale
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.deployments.scale({ /* ... */ });
POST fireworks.inference.v1.accounts.deployments.undelete
POST https://api.fireworks.ai/v1/accounts/{accountId}/deployments/{deploymentId}:undelete
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.deployments.undelete({ /* ... */ });
PATCH fireworks.inference.v1.accounts.deployments.update
PATCH https://api.fireworks.ai/v1/accounts/{accountId}/deployments/{deploymentId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.deployments.update({ /* ... */ });
POST fireworks.inference.v1.accounts.dpoJobs.create
POST https://api.fireworks.ai/v1/accounts/{accountId}/dpoJobs
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.dpoJobs.create({ /* ... */ });
DELETE fireworks.inference.v1.accounts.dpoJobs
DELETE https://api.fireworks.ai/v1/accounts/{accountId}/dpoJobs/{jobId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.dpoJobs.delete({ /* ... */ });
GET fireworks.inference.v1.accounts.dpoJobs
GET https://api.fireworks.ai/v1/accounts/{accountId}/dpoJobs/{jobId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.dpoJobs.get({ /* ... */ });
GET fireworks.inference.v1.accounts.dpoJobs.getMetricsFileEndpoint
GET https://api.fireworks.ai/v1/accounts/{accountId}/dpoJobs/{jobId}:getMetricsFileEndpoint
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.dpoJobs.getMetricsFileEndpoint({ /* ... */ });
GET fireworks.inference.v1.accounts.dpoJobs.list
GET https://api.fireworks.ai/v1/accounts/{accountId}/dpoJobs
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.dpoJobs.list({ /* ... */ });
POST fireworks.inference.v1.accounts.dpoJobs.resume
POST https://api.fireworks.ai/v1/accounts/{accountId}/dpoJobs/{jobId}:resume
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.dpoJobs.resume({ /* ... */ });
POST fireworks.inference.v1.accounts.evaluationJobs.create
POST https://api.fireworks.ai/v1/accounts/{accountId}/evaluationJobs
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.evaluationJobs.create({ /* ... */ });
DELETE fireworks.inference.v1.accounts.evaluationJobs
DELETE https://api.fireworks.ai/v1/accounts/{accountId}/evaluationJobs/{evaluationJobId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.evaluationJobs.delete({ /* ... */ });
GET fireworks.inference.v1.accounts.evaluationJobs
GET https://api.fireworks.ai/v1/accounts/{accountId}/evaluationJobs/{evaluationJobId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.evaluationJobs.get({ /* ... */ });
GET fireworks.inference.v1.accounts.evaluationJobs.getExecutionLogEndpoint
GET https://api.fireworks.ai/v1/accounts/{accountId}/evaluationJobs/{evaluationJobId}:getExecutionLogEndpoint
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.evaluationJobs.getExecutionLogEndpoint({ /* ... */ });
GET fireworks.inference.v1.accounts.evaluationJobs.list
GET https://api.fireworks.ai/v1/accounts/{accountId}/evaluationJobs
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.evaluationJobs.list({ /* ... */ });
POST fireworks.inference.v1.accounts.evaluators.create
POST https://api.fireworks.ai/v1/accounts/{accountId}/evaluatorsV2
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.evaluators.create({ /* ... */ });
DELETE fireworks.inference.v1.accounts.evaluators
DELETE https://api.fireworks.ai/v1/accounts/{accountId}/evaluators/{evaluatorId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.evaluators.delete({ /* ... */ });
GET fireworks.inference.v1.accounts.evaluators
GET https://api.fireworks.ai/v1/accounts/{accountId}/evaluators/{evaluatorId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.evaluators.get({ /* ... */ });
GET fireworks.inference.v1.accounts.evaluators.getBuildLogEndpoint
GET https://api.fireworks.ai/v1/accounts/{accountId}/evaluators/{evaluatorId}:getBuildLogEndpoint
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.evaluators.getBuildLogEndpoint({ /* ... */ });
GET fireworks.inference.v1.accounts.evaluators.getSourceCodeSignedUrl
GET https://api.fireworks.ai/v1/accounts/{accountId}/evaluators/{evaluatorId}:getSourceCodeSignedUrl
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.evaluators.getSourceCodeSignedUrl({ /* ... */ });
POST fireworks.inference.v1.accounts.evaluators.getUploadEndpoint
POST https://api.fireworks.ai/v1/accounts/{accountId}/evaluators/{evaluatorId}:getUploadEndpoint
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.evaluators.getUploadEndpoint({ /* ... */ });
GET fireworks.inference.v1.accounts.evaluators.list
GET https://api.fireworks.ai/v1/accounts/{accountId}/evaluators
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.evaluators.list({ /* ... */ });
PATCH fireworks.inference.v1.accounts.evaluators.update
PATCH https://api.fireworks.ai/v1/accounts/{accountId}/evaluators/{evaluatorId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.evaluators.update({ /* ... */ });
POST fireworks.inference.v1.accounts.evaluators.validateUpload
POST https://api.fireworks.ai/v1/accounts/{accountId}/evaluators/{evaluatorId}:validateUpload
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.evaluators.validateUpload({ /* ... */ });
GET fireworks.inference.v1.accounts
GET https://api.fireworks.ai/v1/accounts/{accountId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.get({ /* ... */ });
GET fireworks.inference.v1.accounts.list
GET https://api.fireworks.ai/v1/accounts
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.list({ /* ... */ });
POST fireworks.inference.v1.accounts.models.create
POST https://api.fireworks.ai/v1/accounts/{accountId}/models
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.models.create({ /* ... */ });
DELETE fireworks.inference.v1.accounts.models
DELETE https://api.fireworks.ai/v1/accounts/{accountId}/models/{modelId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.models.delete({ /* ... */ });
GET fireworks.inference.v1.accounts.models
GET https://api.fireworks.ai/v1/accounts/{accountId}/models/{modelId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.models.get({ /* ... */ });
GET fireworks.inference.v1.accounts.models.getDownloadEndpoint
GET https://api.fireworks.ai/v1/accounts/{accountId}/models/{modelId}:getDownloadEndpoint
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.models.getDownloadEndpoint({ /* ... */ });
POST fireworks.inference.v1.accounts.models.getUploadEndpoint
POST https://api.fireworks.ai/v1/accounts/{accountId}/models/{modelId}:getUploadEndpoint
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.models.getUploadEndpoint({ /* ... */ });
GET fireworks.inference.v1.accounts.models.list
GET https://api.fireworks.ai/v1/accounts/{accountId}/models
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.models.list({ /* ... */ });
POST fireworks.inference.v1.accounts.models.prepare
POST https://api.fireworks.ai/v1/accounts/{accountId}/models/{modelId}:prepare
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.models.prepare({ /* ... */ });
PATCH fireworks.inference.v1.accounts.models.update
PATCH https://api.fireworks.ai/v1/accounts/{accountId}/models/{modelId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.models.update({ /* ... */ });
GET fireworks.inference.v1.accounts.models.validateUpload
GET https://api.fireworks.ai/v1/accounts/{accountId}/models/{modelId}:validateUpload
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.models.validateUpload({ /* ... */ });
POST fireworks.inference.v1.accounts.reinforcementFineTuningJobs.create
POST https://api.fireworks.ai/v1/accounts/{accountId}/reinforcementFineTuningJobs
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.reinforcementFineTuningJobs.create({ /* ... */ });
DELETE fireworks.inference.v1.accounts.reinforcementFineTuningJobs
DELETE https://api.fireworks.ai/v1/accounts/{accountId}/reinforcementFineTuningJobs/{jobId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.reinforcementFineTuningJobs.delete({ /* ... */ });
GET fireworks.inference.v1.accounts.reinforcementFineTuningJobs
GET https://api.fireworks.ai/v1/accounts/{accountId}/reinforcementFineTuningJobs/{jobId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.reinforcementFineTuningJobs.get({ /* ... */ });
GET fireworks.inference.v1.accounts.reinforcementFineTuningJobs.list
GET https://api.fireworks.ai/v1/accounts/{accountId}/reinforcementFineTuningJobs
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.reinforcementFineTuningJobs.list({ /* ... */ });
POST fireworks.inference.v1.accounts.reinforcementFineTuningJobs.resume
POST https://api.fireworks.ai/v1/accounts/{accountId}/reinforcementFineTuningJobs/{jobId}:resume
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.reinforcementFineTuningJobs.resume({ /* ... */ });
POST fireworks.inference.v1.accounts.rlorTrainerJobs.create
POST https://api.fireworks.ai/v1/accounts/{accountId}/rlorTrainerJobs
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.rlorTrainerJobs.create({ /* ... */ });
DELETE fireworks.inference.v1.accounts.rlorTrainerJobs
DELETE https://api.fireworks.ai/v1/accounts/{accountId}/rlorTrainerJobs/{jobId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.rlorTrainerJobs.delete({ /* ... */ });
POST fireworks.inference.v1.accounts.rlorTrainerJobs.executeTrainStep
POST https://api.fireworks.ai/v1/accounts/{accountId}/rlorTrainerJobs/{jobId}:executeTrainStep
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.rlorTrainerJobs.executeTrainStep({ /* ... */ });
GET fireworks.inference.v1.accounts.rlorTrainerJobs
GET https://api.fireworks.ai/v1/accounts/{accountId}/rlorTrainerJobs/{jobId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.rlorTrainerJobs.get({ /* ... */ });
GET fireworks.inference.v1.accounts.rlorTrainerJobs.list
GET https://api.fireworks.ai/v1/accounts/{accountId}/rlorTrainerJobs
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.rlorTrainerJobs.list({ /* ... */ });
POST fireworks.inference.v1.accounts.rlorTrainerJobs.resume
POST https://api.fireworks.ai/v1/accounts/{accountId}/rlorTrainerJobs/{jobId}:resume
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.rlorTrainerJobs.resume({ /* ... */ });
POST fireworks.inference.v1.accounts.secrets.create
POST https://api.fireworks.ai/v1/accounts/{accountId}/secrets
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.secrets.create({ /* ... */ });
DELETE fireworks.inference.v1.accounts.secrets
DELETE https://api.fireworks.ai/v1/accounts/{accountId}/secrets/{secretId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.secrets.delete({ /* ... */ });
GET fireworks.inference.v1.accounts.secrets
GET https://api.fireworks.ai/v1/accounts/{accountId}/secrets/{secretId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.secrets.get({ /* ... */ });
GET fireworks.inference.v1.accounts.secrets.list
GET https://api.fireworks.ai/v1/accounts/{accountId}/secrets
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.secrets.list({ /* ... */ });
PATCH fireworks.inference.v1.accounts.secrets.update
PATCH https://api.fireworks.ai/v1/accounts/{accountId}/secrets/{secretId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.secrets.update({ /* ... */ });
POST fireworks.inference.v1.accounts.supervisedFineTuningJobs.create
POST https://api.fireworks.ai/v1/accounts/{accountId}/supervisedFineTuningJobs
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.supervisedFineTuningJobs.create({ /* ... */ });
DELETE fireworks.inference.v1.accounts.supervisedFineTuningJobs
DELETE https://api.fireworks.ai/v1/accounts/{param}/supervisedFineTuningJobs/{param}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.supervisedFineTuningJobs.delete({ /* ... */ });
GET fireworks.inference.v1.accounts.supervisedFineTuningJobs
GET https://api.fireworks.ai/v1/accounts/{param}/supervisedFineTuningJobs/{param}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.supervisedFineTuningJobs.get({ /* ... */ });
GET fireworks.inference.v1.accounts.supervisedFineTuningJobs.list
GET https://api.fireworks.ai/v1/accounts/{accountId}/supervisedFineTuningJobs
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.supervisedFineTuningJobs.list({ /* ... */ });
POST fireworks.inference.v1.accounts.supervisedFineTuningJobs.resume
POST https://api.fireworks.ai/v1/accounts/{param}/supervisedFineTuningJobs/{param}:resume
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.supervisedFineTuningJobs.resume({ /* ... */ });
POST fireworks.inference.v1.accounts.users.create
POST https://api.fireworks.ai/v1/accounts/{accountId}/users
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.users.create({ /* ... */ });
GET fireworks.inference.v1.accounts.users
GET https://api.fireworks.ai/v1/accounts/{accountId}/users/{userId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.users.get({ /* ... */ });
GET fireworks.inference.v1.accounts.users.list
GET https://api.fireworks.ai/v1/accounts/{accountId}/users
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.users.list({ /* ... */ });
PATCH fireworks.inference.v1.accounts.users.update
PATCH https://api.fireworks.ai/v1/accounts/{accountId}/users/{userId}
成本等级 : expensive
const res = await fireworks.inference.v1.accounts.users.update({ /* ... */ });
GET fireworks.inference.v1.audio.batch
GET https://api.fireworks.ai/v1/accounts/{accountId}/batch_job/{batchId}
成本等级 : expensive
折旧: 烟花公司于2026-06-10日贬值了音频推论,给出了不关闭日期,不再公布其费率,因此成本估算未能关闭. 2026-11-05 (英语)
audio-batch已回复submitted以空请求endpoint_id=audio-prod或者说audio-turbo,而状态端点对其签发的ID进行了404的回答. 真实提交是否运行未证实 。 这种方法仍然发出请求,没有改变,但烟花公司可能会不通知就停止回答.
const res = await fireworks.inference.v1.audio.batch.get({ /* ... */ });
POST fireworks.inference.v1.audio.batch.transcriptions
POST https://api.fireworks.ai/inference/v1/audio/transcriptions
成本等级 : expensive
折旧: 烟花公司于2026-06-10日贬值了音频推论,给出了不关闭日期,不再公布其费率,因此成本估算未能关闭. 2026-11-05 (英语)
audio-batch已回复submitted以空请求endpoint_id=audio-prod或者说audio-turbo,而状态端点对其签发的ID进行了404的回答. 真实提交是否运行未证实 。 这种方法仍然发出请求,没有改变,但烟花公司可能会不通知就停止回答.
const res = await fireworks.inference.v1.audio.batch.transcriptions({ /* ... */ });
POST fireworks.inference.v1.audio.batch.translations
POST https://api.fireworks.ai/inference/v1/audio/translations
成本等级 : expensive
折旧: 烟花公司于2026-06-10日贬值了音频推论,给出了不关闭日期,不再公布其费率,因此成本估算未能关闭. 2026-11-05 (英语)
audio-batch已回复submitted以空请求endpoint_id=audio-prod或者说audio-turbo,而状态端点对其签发的ID进行了404的回答. 真实提交是否运行未证实 。 这种方法仍然发出请求,没有改变,但烟花公司可能会不通知就停止回答.
const res = await fireworks.inference.v1.audio.batch.translations({ /* ... */ });
POST fireworks.inference.v1.audio.transcriptions
POST https://api.fireworks.ai/inference/v1/audio/transcriptions
成本等级 : expensive
折旧: 烟花公司于2026-06-10日贬值了音频推论,给出了不关闭日期,不再公布其费率,因此成本估算未能关闭. 2026-11-05 (英语)
audio-prod在真实路径和编造路径上,401对有效的密钥进行了回答audio-turbo回答503,没有健康的上游。 这种方法仍然发出请求,没有改变,但烟花公司可能会不通知就停止回答.
const res = await fireworks.inference.v1.audio.transcriptions({ /* ... */ });
POST fireworks.inference.v1.audio.translations
POST https://api.fireworks.ai/inference/v1/audio/translations
成本等级 : expensive
折旧: 烟花公司于2026-06-10日贬值了音频推论,给出了不关闭日期,不再公布其费率,因此成本估算未能关闭. 2026-11-05 (英语)
audio-prod在真实路径和编造路径上,401对有效的密钥进行了回答audio-turbo回答503,没有健康的上游。 这种方法仍然发出请求,没有改变,但烟花公司可能会不通知就停止回答.
const res = await fireworks.inference.v1.audio.translations({ /* ... */ });
POST fireworks.inference.v1.chat.completions
POST https://api.fireworks.ai/inference/v1/chat/completions
成本等级 : expensive
const res = await fireworks.inference.v1.chat.completions({ /* ... */ });
POST fireworks.inference.v1.completions
POST https://api.fireworks.ai/inference/v1/completions
成本等级 : expensive
const res = await fireworks.inference.v1.completions({ /* ... */ });
POST fireworks.inference.v1.embeddings
POST https://api.fireworks.ai/inference/v1/embeddings
成本等级 : expensive
const res = await fireworks.inference.v1.embeddings({ /* ... */ });
POST fireworks.inference.v1.rerank
POST https://api.fireworks.ai/inference/v1/rerank
成本等级 : expensive
const res = await fireworks.inference.v1.rerank({ /* ... */ });
POST fireworks.inference.v1.accounts.users.update
POST https://api.fireworks.ai/v1/accounts/{accountId}/users/{userId}
成本等级 : expensive
const res = await fireworks.post.inference.v1.accounts.users.update({ /* ... */ });
POST fireworks.inference.v1.messages
POST https://api.fireworks.ai/inference/v1/messages
成本等级 : expensive
const res = await fireworks.post.inference.v1.messages({ /* ... */ });
POST fireworks.post.stream.inference.v1.messages
POST https://api.fireworks.ai/inference/v1/messages
成本等级 : expensive
const res = await fireworks.post.stream.inference.v1.messages({ /* ... */ });
中间软件
import { createFireworks, withRetry } from "@apicity/fireworks";
const fireworks = createFireworks({ apiKey: process.env.FIREWORKS_API_KEY! });
const listAccounts = withRetry(fireworks.inference.v1.accounts.list, { retries: 3 });
部分 辣点 单相机.
许可证
麻省理工学院——见 自由主义。 。 。 .