mirror of
https://github.com/norandom/OpenUltraSAST.git
synced 2026-10-18 04:29:27 +00:00
feat(learning-harness): chat endpoint independent of embeddings, with the DeepSeek adapter
Task 1.2. Detectors, the classifier's model tier, the proposer and the judge resolve one chat endpoint: an explicit override first, then the scripted client flag, then an endpoint named in the config, then DeepSeek, then OpenRouter. Prove-order retrieval keeps its own embedding endpoint untouched, because the chat provider we default to offers none. The adapter handles the provider's documented quirks: thinking is on by default there and silently ignores temperature, so it is disabled explicitly; the reasoning field is kept on the reply for the caller to replay across tool turns; JSON mode has no schema and occasionally returns empty content, so one retry is built in, never a loop; token log-probabilities are requested only when the caller wants them, and their mean is what lets the classifier abstain. Cost comes from the provider's own usage fields, cache hits priced separately, and a model with no recorded price reports no cost rather than a made-up one. The endpoint record deliberately holds no key, so it is safe to journal. Co-Authored-By: Claude Opus 5 (1M context) <[email protected]> Claude-Session: https://claude.ai/code/session_01T7XD7x5KonMdmGRkp63EQG
This commit is contained in:
Marius Ciepluch
co-authored by
Claude Opus 5
Claude Opus 5
parent
289648338c
commit
944df86d12
@@ -0,0 +1,215 @@
|
||||
"""The chat endpoint, resolved independently of embeddings (learning-harness, Req 6.6).
|
||||
|
||||
Detectors, the classifier's model tier, the proposer and the judge all talk to one chat endpoint;
|
||||
prove-order retrieval keeps its own embedding endpoint, because the chat provider we default to has
|
||||
none. Resolution order is explicit override, the scripted client flag, an endpoint named in the
|
||||
config, DeepSeek, then OpenRouter. ``DeepSeekChatClient`` adapts the provider's quirks: thinking is
|
||||
on by default there and silently ignores ``temperature``, so it is disabled explicitly; the reasoning
|
||||
field is kept on the reply for the caller to replay across tool turns; JSON mode has no schema and
|
||||
occasionally returns empty content, so one retry is built in. Prices are data, and an unpriced model
|
||||
reports no cost rather than a made-up one.
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import os
|
||||
from collections.abc import Mapping, Sequence
|
||||
from dataclasses import dataclass
|
||||
from typing import Literal
|
||||
|
||||
from ..config import ResolvedConfig
|
||||
from ..provider.openrouter import OpenRouterChatClient, OpenRouterError
|
||||
from ..tool_hunter import CLIENT_ENV, ChatClient, ChatResponse, chat_response_from_message, scripted_hunter_client
|
||||
|
||||
DEEPSEEK_BASE_URL = "https://api.deepseek.com" # no version suffix: the platform rejects /v1
|
||||
DEEPSEEK_KEY_ENV = "DEEPSEEK_API_KEY"
|
||||
DEEPSEEK_BASE_ENV = "DEEPSEEK_BASE_URL"
|
||||
OPENROUTER_BASE_URL = "https://openrouter.ai/api/v1"
|
||||
DEFAULT_DETECTOR_MODEL = "deepseek-v4-flash"
|
||||
DEFAULT_JUDGE_MODEL = "deepseek-v4-pro"
|
||||
Provider = Literal["deepseek", "openrouter", "custom", "scripted"]
|
||||
_SCRIPTED_FLAGS = frozenset({"scripted", "script", "dump", "dump-only", "unsafe", "unsafe-snippet"})
|
||||
|
||||
|
||||
@dataclass(frozen=True)
|
||||
class Prices:
|
||||
"""USD per million tokens. DeepSeek platform pricing, peak rate, recorded 2026-09-06."""
|
||||
|
||||
cache_hit_per_m: float
|
||||
input_per_m: float
|
||||
output_per_m: float
|
||||
|
||||
|
||||
_PRICES: dict[str, Prices] = {
|
||||
"deepseek-v4-flash": Prices(cache_hit_per_m=0.014, input_per_m=0.44, output_per_m=1.32),
|
||||
"deepseek-v4-flash-vision-exp": Prices(cache_hit_per_m=0.014, input_per_m=0.44, output_per_m=1.32),
|
||||
"deepseek-v4-pro": Prices(cache_hit_per_m=0.044, input_per_m=1.32, output_per_m=3.96),
|
||||
}
|
||||
|
||||
|
||||
def price_of(model: str) -> Prices | None:
|
||||
return _PRICES.get(model.strip())
|
||||
|
||||
|
||||
@dataclass(frozen=True)
|
||||
class ChatEndpoint:
|
||||
"""What the run should record about where the model calls went. Never holds the key."""
|
||||
|
||||
provider: Provider
|
||||
base_url: str
|
||||
thinking: bool # the provider's models reason by default, so the adapter disables it explicitly
|
||||
prices: Prices | None = None
|
||||
|
||||
def cost(self, usage: Mapping[str, object] | None) -> float | None:
|
||||
"""Dollars for one call from the provider's own usage fields, or None when the model has no recorded price."""
|
||||
if self.prices is None or not isinstance(usage, Mapping):
|
||||
return None
|
||||
cache_hit = _tokens(usage.get("prompt_cache_hit_tokens"))
|
||||
miss = usage.get("prompt_cache_miss_tokens")
|
||||
input_tokens = _tokens(miss) if miss is not None else max(_tokens(usage.get("prompt_tokens")) - cache_hit, 0)
|
||||
output_tokens = _tokens(usage.get("completion_tokens"))
|
||||
return (
|
||||
cache_hit * self.prices.cache_hit_per_m + input_tokens * self.prices.input_per_m + output_tokens * self.prices.output_per_m
|
||||
) / 1_000_000
|
||||
|
||||
|
||||
class DeepSeekChatClient:
|
||||
"""Adapt an OpenAI-shaped chat client to what detectors and the classifier need."""
|
||||
|
||||
def __init__(self, client: OpenRouterChatClient, *, disable_thinking: bool = True, endpoint: ChatEndpoint | None = None) -> None:
|
||||
self._client = client
|
||||
self._disable_thinking = disable_thinking
|
||||
self.endpoint = endpoint
|
||||
self.usage: list[dict[str, object]] = [] # one entry per call, for the round's cost meter
|
||||
|
||||
def complete(
|
||||
self,
|
||||
*,
|
||||
model: str,
|
||||
messages: list[dict[str, object]],
|
||||
tools: list[dict[str, object]],
|
||||
timeout_seconds: int = 60,
|
||||
json_object: bool = False,
|
||||
logprobs: bool = False,
|
||||
) -> ChatResponse:
|
||||
response = self._call(
|
||||
model=model, messages=messages, tools=tools, timeout_seconds=timeout_seconds, json_object=json_object, logprobs=logprobs
|
||||
)
|
||||
if json_object and not (response.content or "").strip():
|
||||
response = self._call(
|
||||
model=model, messages=messages, tools=tools, timeout_seconds=timeout_seconds, json_object=json_object, logprobs=logprobs
|
||||
) # documented behaviour: JSON mode occasionally returns empty content. One retry, never a loop.
|
||||
return response
|
||||
|
||||
def cost_usd(self) -> float:
|
||||
"""Recorded cost of the calls made so far; calls whose model has no price contribute nothing."""
|
||||
if self.endpoint is None:
|
||||
return 0.0
|
||||
return sum(self.endpoint.cost(entry) or 0.0 for entry in self.usage)
|
||||
|
||||
def _call(
|
||||
self,
|
||||
*,
|
||||
model: str,
|
||||
messages: Sequence[Mapping[str, object]],
|
||||
tools: Sequence[Mapping[str, object]],
|
||||
timeout_seconds: int,
|
||||
json_object: bool,
|
||||
logprobs: bool,
|
||||
) -> ChatResponse:
|
||||
extra: dict[str, object] = {}
|
||||
if self._disable_thinking:
|
||||
extra["thinking"] = {"type": "disabled"} # otherwise temperature is silently ignored
|
||||
if json_object:
|
||||
extra["response_format"] = {"type": "json_object"}
|
||||
if logprobs:
|
||||
extra["logprobs"] = True
|
||||
extra["top_logprobs"] = 1
|
||||
raw = getattr(self._client, "complete_chat_raw", None)
|
||||
if callable(raw):
|
||||
payload = raw(
|
||||
model=model, messages=list(messages), tools=list(tools) or None, timeout_seconds=timeout_seconds, extra_body=extra
|
||||
)
|
||||
usage = payload.get("usage") if isinstance(payload, dict) else None
|
||||
if isinstance(usage, dict):
|
||||
self.usage.append(dict(usage))
|
||||
message = _message_of(payload)
|
||||
else: # a stub or a client without the raw call
|
||||
message = self._client.complete_chat(
|
||||
model=model, messages=list(messages), tools=list(tools) or None, timeout_seconds=timeout_seconds, extra_body=extra
|
||||
)
|
||||
return chat_response_from_message(message)
|
||||
|
||||
|
||||
def resolve_chat_endpoint(config: ResolvedConfig, *, override: ChatClient | None = None) -> tuple[ChatClient, ChatEndpoint] | None:
|
||||
"""The chat client for detectors, classifier, proposer and judge, with what to record about it."""
|
||||
if override is not None:
|
||||
return override, ChatEndpoint(provider="scripted", base_url="", thinking=False)
|
||||
flag = os.environ.get(CLIENT_ENV, "").strip().lower()
|
||||
if flag in _SCRIPTED_FLAGS:
|
||||
return scripted_hunter_client(flag), ChatEndpoint(provider="scripted", base_url="", thinking=False)
|
||||
model = resolve_models(config)[0]
|
||||
configured = config.models.chat_base_url
|
||||
key_env = config.models.chat_api_key_env
|
||||
if configured or key_env:
|
||||
key = os.environ.get(key_env or DEEPSEEK_KEY_ENV, "")
|
||||
base_url = configured or DEEPSEEK_BASE_URL
|
||||
return _deepseek_client(key, base_url, model, provider=_provider_of(base_url))
|
||||
deepseek_key = os.environ.get(DEEPSEEK_KEY_ENV)
|
||||
if deepseek_key:
|
||||
return _deepseek_client(deepseek_key, os.environ.get(DEEPSEEK_BASE_ENV, DEEPSEEK_BASE_URL), model, provider="deepseek")
|
||||
if os.environ.get("OPENROUTER_API_KEY"):
|
||||
try:
|
||||
client = OpenRouterChatClient.from_env()
|
||||
except OpenRouterError:
|
||||
return None
|
||||
endpoint = ChatEndpoint(provider="openrouter", base_url=client.base_url, thinking=False, prices=price_of(model))
|
||||
return DeepSeekChatClient(client, disable_thinking=False, endpoint=endpoint), endpoint
|
||||
return None
|
||||
|
||||
|
||||
def resolve_models(config: ResolvedConfig) -> tuple[str, str]:
|
||||
"""(detector model, judge model). `[learning]` wins, then `[models]`, then the shipped defaults."""
|
||||
detector = config.learning.detector_model or config.models.hunter or DEFAULT_DETECTOR_MODEL
|
||||
judge = config.learning.judge_model or config.models.judge or DEFAULT_JUDGE_MODEL
|
||||
return detector, judge
|
||||
|
||||
|
||||
def _deepseek_client(key: str, base_url: str, model: str, *, provider: Provider) -> tuple[ChatClient, ChatEndpoint]:
|
||||
endpoint = ChatEndpoint(provider=provider, base_url=base_url.rstrip("/"), thinking=True, prices=price_of(model))
|
||||
client = OpenRouterChatClient(api_key=key, base_url=endpoint.base_url)
|
||||
return DeepSeekChatClient(client, endpoint=endpoint), endpoint
|
||||
|
||||
|
||||
def _provider_of(base_url: str) -> Provider:
|
||||
host = base_url.lower()
|
||||
if "deepseek" in host:
|
||||
return "deepseek"
|
||||
return "openrouter" if "openrouter" in host else "custom"
|
||||
|
||||
|
||||
def _message_of(payload: object) -> Mapping[str, object]:
|
||||
if isinstance(payload, Mapping):
|
||||
choices = payload.get("choices")
|
||||
if isinstance(choices, Sequence) and choices and isinstance(choices[0], Mapping):
|
||||
message = choices[0].get("message")
|
||||
if isinstance(message, Mapping):
|
||||
return message
|
||||
return {}
|
||||
|
||||
|
||||
def _tokens(value: object) -> int:
|
||||
return int(value) if isinstance(value, int | float) else 0
|
||||
|
||||
|
||||
__all__ = [
|
||||
"DEEPSEEK_BASE_URL",
|
||||
"DEFAULT_DETECTOR_MODEL",
|
||||
"DEFAULT_JUDGE_MODEL",
|
||||
"ChatEndpoint",
|
||||
"DeepSeekChatClient",
|
||||
"Prices",
|
||||
"price_of",
|
||||
"resolve_chat_endpoint",
|
||||
"resolve_models",
|
||||
]
|
||||
Reference in New Issue
Block a user