unhardcoded
開源 · AI 模型的策略路由層

建構 AI 系統，無需寫死模型選擇。

每個請求附帶一條策略。unhardcoded 篩選即時模型目錄，挑選符合規則、成本最低的候選模型，透過您自己的供應商金鑰執行，並回傳決策的追蹤紀錄。

request
Prompt + policy
tools · intel ge 0.5 · cheapest
→
candidateprice_outintelverdict
deepseek-v4-flash$0.400.465未達下限
minimax-m2.7$0.500.496未達下限
deepseek-v4-pro$1.500.515勝出
glm-5.1$2.000.514通過
gpt-5.5$10.000.602通過
→
trace
deepseek-v4-pro
cheapest passing · 412 ms
fp 301140696-1054914287
路由單次呼叫
將 SDK 指向託管位址，附加策略，讀取追蹤紀錄。
快速入門 →
執行工作流程
將路由呼叫組合成 DAG：分類、扇出、評判、合併。
查看模式 →
讀取追蹤紀錄
每次決策都留下憑證：哪些候選模型被考慮、誰勝出，以及原因。
檢視決策 →
核心概念

模型路由迴圈

一條策略決定一次呼叫。相同輸入、相同目錄、相同決策。每次皆然，且有書面紀錄。

Request
您的呼叫送達相容 OpenAI 的端點，並攜帶一條策略。
Policy sigma-pol/v2
一個小型項(term)：filter、rank、select、mutate、fallback。
Catalog
含有價格、基準分數與能力的（供應商、模型）候選模型即時集合。
Filter
排除不符合規則的候選模型，不存在靜默降級。
Rank
對通過篩選的候選模型評分：最便宜、能力最強、速度最快，由您決定。
Select
取排名最高的一個，或使用 top_k 備援串接。
Run
透過您的供應商金鑰執行推論，發生錯誤時自動備援。
Trace
記錄所有候選模型、勝出者及指紋(fingerprint)的憑證。

深入了解策略運作原理 →

組合路由呼叫

工作流程模式

工作流程是由路由步驟組成的有界無環圖。每個節點攜帶自己的策略並獨立路由；整個圖產生一條統一的追蹤紀錄。這正是 unhardcoded 從「帶規則的路由」演變為完整系統的地方。

支援工單linear · guard

低成本分類，依品質下限起草回覆，最後由一個強力 no-log 守衛在送出前決定是否拒絕。

關鍵：最後一步是由策略強制執行的閘門，可以中止送出，而不只是寄望它能發揮作用。

工單
→
triageclassify · extract iddeepseek-v4-flash
→
draftreply · quality floorgemini-3.1-pro-preview
→
guardbrand · PII · no-loggpt-5.5 · can abort
→
回覆
查看 flow_ir
flow.support-ticket.json
["flow", {
  "u": {"kind": "input"},
  "t": {"kind": "llm", "system": "Classify the ticket and extract the account id as JSON.",
    "policy": ["policy", ["and", ["meets_req"], ["not", ["is", "disabled"]], ["has_cap", "supports_json_mode"]],
      ["neg", ["normalize", ["field", "price_out"]]], ["argmax"], ["id"], ["always", {"action": "next_candidate"}]],
    "inputs": ["u"]},
  "d": {"kind": "llm", "system": "Write a reply using the ticket and the triage.",
    "policy": ["policy", ["and", ["meets_req"], ["not", ["is", "disabled"]], ["cmp", "bench_intelligence", "ge", 0.55]],
      ["neg", ["normalize", ["field", "price_out"]]], ["argmax"], ["id"], ["always", {"action": "next_candidate"}]],
    "inputs": ["u", "t"], "template": "Ticket:\n$1\n\nTriage:\n$2"},
  "g": {"kind": "llm", "system": "Check brand voice, PII, refund limits. Refuse if any fail.",
    "policy": ["policy", ["and", ["meets_req"], ["not", ["is", "disabled"]], ["is", "no_log"]],
      ["field", "bench_intelligence"], ["argmax"], ["id"], ["always", {"action": "next_candidate"}]],
    "inputs": ["d"]},
  "out": {"kind": "output", "inputs": ["g"]}
}]
起草 → 批評 → 修改linear · refine

先由低成本模型生成初稿，再由強力批評者列出缺陷，最後重寫並修正每個問題。

關鍵：在預算範圍內實現品質跨越——大部分權杖消耗在低成本模型上。

問題
→
draftcheapestdeepseek-v4-flash
→
critiquelist the flawsgpt-5.5
→
revisefix every point · fan-ingpt-5.5
→
答案
查看 flow_ir
flow.draft-critique-revise.json
["flow", {
  "u": {"kind": "input"},
  "d": {"kind": "llm", "system": "Draft an answer.",
    "policy": ["policy", ["and", ["meets_req"], ["not", ["is", "disabled"]]],
      ["neg", ["normalize", ["field", "price_out"]]], ["argmax"], ["id"], ["always", {"action": "next_candidate"}]],
    "inputs": ["u"]},
  "c": {"kind": "llm", "system": "Critique the draft: list concrete flaws and gaps.",
    "policy": ["policy", ["and", ["meets_req"], ["not", ["is", "disabled"]]],
      ["field", "bench_intelligence"], ["argmax"], ["id"], ["always", {"action": "next_candidate"}]],
    "inputs": ["d"]},
  "r": {"kind": "llm", "system": "Rewrite the answer, fixing every point in the critique.",
    "policy": ["policy", ["and", ["meets_req"], ["not", ["is", "disabled"]]],
      ["field", "bench_intelligence"], ["argmax"], ["id"], ["always", {"action": "next_candidate"}]],
    "inputs": ["u", "d", "c"], "template": "Question:\n$1\n\nDraft:\n$2\n\nCritique:\n$3"},
  "out": {"kind": "output", "inputs": ["r"]}
}]
N 選一 + 評判ensemble

以同一強力策略進行 N 次種子採樣，再由評判者從中擇優選出勝出結果。

關鍵：sample 以確定性方式在排名靠前的候選模型中分散採樣——可重現的多樣性，而非隨機。

提示詞
→
draft Asample · T=0.5gpt-5.5
draft Bsample · T=0.5gpt-5.4
draft Csample · T=0.5gemini-3.1-pro-preview
→
judgerank · pick winnergpt-5.5
→
答案
查看 flow_ir
flow.best-of-n.json
["flow", {
  "u": {"kind": "input"},
  "n1": {"kind": "llm", "system": "Answer the question.",
    "policy": ["policy", ["and", ["meets_req"], ["not", ["is", "disabled"]]],
      ["field", "bench_intelligence"], ["sample", 0.5], ["id"], ["always", {"action": "next_candidate"}]],
    "inputs": ["u"]},
  // n2, n3: the same policy, two more seeded draws
  "j": {"kind": "llm", "system": "Pick the single best candidate; return it verbatim.",
    "policy": ["policy", ["and", ["meets_req"], ["not", ["is", "disabled"]]],
      ["field", "bench_intelligence"], ["argmax"], ["id"], ["always", {"action": "next_candidate"}]],
    "inputs": ["n1", "n2", "n3"], "template": "A:\n$1\n\nB:\n$2\n\nC:\n$3"},
  "out": {"kind": "output", "inputs": ["j"]}
}]
模型評審團fan-out · fuse

三個依族系固定的模型並行起草，第四個模型將它們合成為一個更優的單一答案。

關鍵：family_eq 鎖定確切的模型系列，使評審團可重現——而非「今天最便宜的那個」。

問題
→
draft afamily_eqgemini-3.1-pro-preview
draft bfamily_eqclaude-opus-4-8
draft cfamily_eqdeepseek-v4-flash
→
fusesynthesize · fan-ingpt-5.5
→
答案
查看 flow_ir
flow.panel.json
["flow", {
  "u": {"kind": "input"},
  "a": {"kind": "llm", "system": "Draft an answer.",
    "policy": ["policy", ["and", ["meets_req"], ["not", ["is", "disabled"]], ["family_eq", "gemini-3.1-pro-preview"]],
      ["zero"], ["argmax"], ["id"], ["always", {"action": "next_candidate"}]], "inputs": ["u"]},
  // b: family_eq "claude-opus-4-8"  ·  c: family_eq "deepseek-v4-flash"
  "f": {"kind": "llm", "system": "Synthesize the single best answer from the drafts.",
    "policy": ["policy", ["and", ["meets_req"], ["not", ["is", "disabled"]]],
      ["field", "bench_intelligence"], ["argmax"], ["id"], ["always", {"action": "next_candidate"}]],
    "inputs": ["a", "b", "c"]},
  "out": {"kind": "output", "inputs": ["f"]}
}]
專家分工fan-out · merge

推論器與程式碼產生器在不同篩選條件下並行執行，再由合併步驟整合。

關鍵：每個分支為其任務（推論 vs 程式設計）挑選最合適的模型，而非讓單一模型包辦所有工作。

問題
→
reasonis cap_reasoninggpt-5.5
codecoding top-5gpt-5.4
→
mergeone answer · fan-ingemini-3.1-pro-preview
→
答案
查看 flow_ir
flow.specialist-split.json
["flow", {
  "u": {"kind": "input"},
  "rz": {"kind": "llm", "system": "Reason through the problem step by step.",
    "policy": ["policy", ["and", ["meets_req"], ["not", ["is", "disabled"]], ["is", "cap_reasoning"]],
      ["field", "bench_intelligence"], ["argmax"], ["id"], ["always", {"action": "next_candidate"}]], "inputs": ["u"]},
  "cd": {"kind": "llm", "system": "Produce any code the problem needs.",
    "policy": ["policy", ["and", ["meets_req"], ["not", ["is", "disabled"]], ["cmp", "bench_coding_rank", "le", 5]],
      ["field", "bench_coding"], ["argmax"], ["id"], ["always", {"action": "next_candidate"}]], "inputs": ["u"]},
  "m": {"kind": "llm", "system": "Merge the reasoning and the code into one answer.",
    "policy": ["policy", ["and", ["meets_req"], ["not", ["is", "disabled"]]],
      ["field", "bench_intelligence"], ["argmax"], ["id"], ["always", {"action": "next_candidate"}]],
    "inputs": ["rz", "cd"], "template": "Reasoning:\n$1\n\nCode:\n$2"},
  "out": {"kind": "output", "inputs": ["m"]}
}]

完整工作流程指南：節點類型、圖限制、範本 →

送出初始請求

快速入門

unhardcoded 相容 OpenAI 介面。相比一般呼叫，僅需三處變更。

  1. 將 SDK 指向託管位址。
    修改 baseURL；SDK 的其餘部分維持不變。
    client.ts
    const client = new OpenAI({
  baseURL: "https://<your-host>/v1",
  apiKey: process.env.UNHARDCODED_KEY,
});
  2. 在呼叫中附加策略。
    在後端建構 policy_ir，與 messages 一同傳送。路由來自策略，因此 model 只是追蹤紀錄標籤。
    route.ts
    const res = await client.chat.completions.create({
  model: "policy:support",
  policy_ir: ["policy",
    ["and", ["meets_req"], ["not", ["is", "disabled"]],
           ["cmp", "bench_intelligence", "ge", 0.5]],   // filter
    ["neg", ["normalize", ["field", "price_out"]]],          // rank: cheapest
    ["argmax"], ["id"], ["always", {"action": "next_candidate"}]],
  messages,
});
  3. 讀取追蹤紀錄。
    回應中包含決策內容：所選模型、經過排序和拒絕的候選模型，以及策略指紋(fingerprint)。
    response · trace (illustrative)
    {
  "chosen": { "model_family": "deepseek-v4-pro", "price_out": 1.5 },
  "trace": {
    "policy_fingerprint": "301140696-1054914287",
    "rejected": [{ "model_family": "deepseek-v4-flash", "reason": "cmp bench_intelligence ge 0.5" }],
    "total_latency_ms": 425
  }
}

完整快速入門：驗證、試執行、可執行範例 →

複製一個起點

策略預設

以卡片形式呈現的常用路由模式。閱讀規則，複製策略，調整下限與上限。

低成本適當品質
在不低於品質下限的前提下降低成本。
filtertools-met · bench_intelligence ge 0.5
rankcheapest price_out
fallback下一個通過的候選模型
查看 JSON
cheapest-decent.json
["policy", ["and", ["meets_req"], ["not", ["is", "disabled"]],
         ["cmp", "bench_intelligence", "ge", 0.5]],
  ["neg", ["normalize", ["field", "price_out"]]],
  ["argmax"], ["id"], ["always", {"action": "next_candidate"}]]
智慧均衡
無強烈偏好：綜合權衡能力與價格。
filtertools-met · not disabled
rank0.6 intelligence + 0.4 cheap
fallback下一個通過的候選模型
查看 JSON
smart-balance.json
["policy", ["and", ["meets_req"], ["not", ["is", "disabled"]]],
  ["add",
    ["scale", 0.6, ["normalize", ["field", "bench_intelligence"]]],
    ["scale", 0.4, ["neg", ["normalize", ["field", "price_out"]]]]],
  ["argmax"], ["id"], ["always", {"action": "next_candidate"}]]
優先智慧能力
關鍵任務，無成本上限：選擇能力更勝一籌的模型。
filtertools-met · not disabled
rankhighest bench_intelligence
fallback下一個通過的候選模型
查看 JSON
best-intelligence.json
["policy", ["and", ["meets_req"], ["not", ["is", "disabled"]]],
  ["field", "bench_intelligence"],
  ["argmax"], ["id"], ["always", {"action": "next_candidate"}]]
僅限推論模型
任務需要具備推論能力的模型。
filter+ is cap_reasoning
rankhighest intelligence
fallback下一個通過的候選模型
查看 JSON
reasoning-only.json
["policy", ["and", ["meets_req"], ["not", ["is", "disabled"]], ["is", "cap_reasoning"]],
  ["field", "bench_intelligence"],
  ["argmax"], ["id"], ["always", {"action": "next_candidate"}]]
視覺 · 低成本
圖像輸入：能夠處理圖像且成本最低的模型。
filter+ is in_image
rankcheapest price_out
fallback下一個通過的候選模型
查看 JSON
vision-cheapest.json
["policy", ["and", ["meets_req"], ["not", ["is", "disabled"]], ["is", "in_image"]],
  ["neg", ["normalize", ["field", "price_out"]]],
  ["argmax"], ["id"], ["always", {"action": "next_candidate"}]]
長情境 RAG
需要大型情境視窗時，選擇能滿足要求且成本最低的模型。
filter+ context ge 200000
rankcheapest price_out
fallback下一個通過的候選模型
查看 JSON
long-context-rag.json
["policy", ["and", ["meets_req"], ["not", ["is", "disabled"]], ["cmp", "context", "ge", 200000]],
  ["neg", ["normalize", ["field", "price_out"]]],
  ["argmax"], ["id"], ["always", {"action": "next_candidate"}]]
代理人調度
為代理人迴圈選擇工具呼叫能力出色的模型。
filter+ tools · bench_agentic_rank le 5
rankhighest bench_agentic
fallback下一個通過的候選模型
查看 JSON
agentic-fleet.json
["policy", ["and", ["meets_req"], ["not", ["is", "disabled"]],
         ["has_cap", "supports_tools"], ["cmp", "bench_agentic_rank", "le", 5]],
  ["field", "bench_agentic"],
  ["argmax"], ["id"], ["always", {"action": "next_candidate"}]]
私密 / 合規
僅限 TEE 且禁止紀錄：能力優先。
filter+ is has_tee · is no_log
rankhighest intelligence
fallback下一個通過的候選模型
查看 JSON
private-compliant.json
["policy", ["and", ["meets_req"], ["not", ["is", "disabled"]],
         ["is", "has_tee"], ["is", "no_log"]],
  ["field", "bench_intelligence"],
  ["argmax"], ["id"], ["always", {"action": "next_candidate"}]]

所有預設及實用配方 →

信任憑證

每次決策都留下憑證

追蹤紀錄是模型選擇過程的結構化、可重播記錄：哪些模型被納入考量、每個模型為何通過或遭拒、執行了什麼，以及如何重現。

decision_trace
deepseek-v4-flashrejected · cmp bench_intelligence ge 0.5
minimax-m2.7rejected · cmp bench_intelligence ge 0.5
deepseek-v4-proranked #1 · chosen
glm-5.1ranked #2 · cascade
gpt-5.5ranked #3 · cascade
deepseek-v4-prodecision_path · attempted · 412 ms
policy_fingerprint 301140696-1054914287 · sigma-pol/v2
rejected[]
所有被篩除的候選模型，以及導致其遭拒的確切規則。
ranked[]
依選擇順序排列的通過候選模型；排名最高者為選中的模型，其餘為備援串接。
decision_path[]
嘗試紀錄；每次備援跳轉都記錄在此。
policy_fingerprint
正規化策略的穩定 ID；識別碼由託管端的 sha256 計算。工作流程還包含 flow_fingerprint。

重播與除錯的運作原理 →

需要細節時

參考文件

完整規格，獨立於閱讀路徑之外。

API 參考
端點、驗證、請求結構
policy_ir
語法與五個插槽
flow_ir
節點類型、圖限制、範本
運算子 & 欄位
簽名與目錄詞彙表
追蹤紀錄 & 重播
追蹤結構、合規性
策略
概念、預設、配方
自行託管
託管服務、目錄、供應商金鑰
疑難排解
錯誤、無候選模型、備援

GitHub 原始碼：unhardcoded（託管服務）· unhardcoded-engine（策略引擎）