Adriano/Multi_Swarm_Coevolutive
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat(llm): full multi-tier S/A/B/C/D with routing + pricing

Browse Source 
Estende ModelTier a 5 livelli (S/A/B/C/D) con routing automatico:
S/A/B via Anthropic SDK, C/D via OpenRouter (OpenAI SDK). Aggiunge
prezzi per tier S (Opus), A (Sonnet placeholder) e D (Llama). Refactor
LLMClient.complete con dispatch tramite tier_models map e helper
_call_anthropic / _call_openrouter. Settings esposte per tutti e 5
i modelli env-configurabili.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
AdrianoDev
2026-05-10 09:18:57 +02:00
parent 7482600146
commit 33d8e275e7
10 changed files with 241 additions and 36 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/multi_swarm/config.py
+4 -1
View File
@@ -26,8 +26,11 @@ class Settings(BaseSettings):
openrouter_api_key: SecretStr
anthropic_api_key: SecretStr | None = None
llm_model_tier_c: str = "qwen/qwen-2.5-72b-instruct"
llm_model_tier_s: str = "claude-opus-4-7"
llm_model_tier_a: str = "claude-sonnet-4-6"
llm_model_tier_b: str = "claude-sonnet-4-6"
llm_model_tier_c: str = "qwen/qwen-2.5-72b-instruct"
llm_model_tier_d: str = "meta-llama/llama-3.3-70b-instruct"
openrouter_base_url: str = "https://openrouter.ai/api/v1"
run_name: str = "phase1-spike-001"
src/multi_swarm/genome/hypothesis.py
+3
View File
@@ -8,8 +8,11 @@ from typing import Any
class ModelTier(StrEnum):
S = "S" # top-tier reasoning (Opus / equivalent) via Anthropic
A = "A" # premium override via Anthropic
B = "B" # Sonnet 4.6 via Anthropic
C = "C" # Qwen 2.5 72B via OpenRouter
D = "D" # ultra-economic (Llama / cheap models) via OpenRouter
@dataclass
src/multi_swarm/llm/client.py
+68 -28
View File
@@ -16,8 +16,11 @@ from tenacity import (
from ..genome.hypothesis import HypothesisAgentGenome, ModelTier
# Modelli configurati per Phase 1
MODEL_TIER_C = "qwen/qwen-2.5-72b-instruct" # via OpenRouter
MODEL_TIER_S = "claude-opus-4-7" # via Anthropic
MODEL_TIER_A = "claude-sonnet-4-6" # via Anthropic (premium override)
MODEL_TIER_B = "claude-sonnet-4-6" # via Anthropic
MODEL_TIER_C = "qwen/qwen-2.5-72b-instruct" # via OpenRouter
MODEL_TIER_D = "meta-llama/llama-3.3-70b-instruct" # via OpenRouter
OPENROUTER_BASE_URL = "https://openrouter.ai/api/v1"
# Errori transient: retry. RateLimit/Auth/InvalidRequest: NO retry.
@@ -41,17 +44,33 @@ class CompletionResult:
class LLMClient:
_ANTHROPIC_TIERS: tuple[ModelTier, ...] = (ModelTier.S, ModelTier.A, ModelTier.B)
_OPENROUTER_TIERS: tuple[ModelTier, ...] = (ModelTier.C, ModelTier.D)
def __init__(
self,
openrouter_api_key: str,
anthropic_api_key: str | None = None,
model_tier_c: str = MODEL_TIER_C,
model_tier_s: str = MODEL_TIER_S,
model_tier_a: str = MODEL_TIER_A,
model_tier_b: str = MODEL_TIER_B,
model_tier_c: str = MODEL_TIER_C,
model_tier_d: str = MODEL_TIER_D,
openrouter_base_url: str = OPENROUTER_BASE_URL,
) -> None:
self.model_tier_c = model_tier_c
self.model_tier_s = model_tier_s
self.model_tier_a = model_tier_a
self.model_tier_b = model_tier_b
self.model_tier_c = model_tier_c
self.model_tier_d = model_tier_d
self.openrouter_base_url = openrouter_base_url
self._tier_models: dict[ModelTier, str] = {
ModelTier.S: model_tier_s,
ModelTier.A: model_tier_a,
ModelTier.B: model_tier_b,
ModelTier.C: model_tier_c,
ModelTier.D: model_tier_d,
}
self._openrouter = OpenAI(api_key=openrouter_api_key, base_url=openrouter_base_url)
self._anthropic = Anthropic(api_key=anthropic_api_key) if anthropic_api_key else None
@@ -68,32 +87,53 @@ class LLMClient:
user: str,
max_tokens: int = 2000,
) -> CompletionResult:
if genome.model_tier == ModelTier.C:
resp = self._openrouter.chat.completions.create(
model=self.model_tier_c,
messages=[
{"role": "system", "content": system},
{"role": "user", "content": user},
],
temperature=genome.temperature,
top_p=genome.top_p,
max_tokens=max_tokens,
)
usage = resp.usage
assert usage is not None
return CompletionResult(
text=resp.choices[0].message.content or "",
input_tokens=usage.prompt_tokens,
output_tokens=usage.completion_tokens,
tier=ModelTier.C,
model=self.model_tier_c,
)
model = self._tier_models[genome.model_tier]
if genome.model_tier in self._ANTHROPIC_TIERS:
return self._call_anthropic(genome, system, user, max_tokens, model)
return self._call_openrouter(genome, system, user, max_tokens, model)
def _call_openrouter(
self,
genome: HypothesisAgentGenome,
system: str,
user: str,
max_tokens: int,
model: str,
) -> CompletionResult:
resp = self._openrouter.chat.completions.create(
model=model,
messages=[
{"role": "system", "content": system},
{"role": "user", "content": user},
],
temperature=genome.temperature,
top_p=genome.top_p,
max_tokens=max_tokens,
)
usage = resp.usage
assert usage is not None
return CompletionResult(
text=resp.choices[0].message.content or "",
input_tokens=usage.prompt_tokens,
output_tokens=usage.completion_tokens,
tier=genome.model_tier,
model=model,
)
def _call_anthropic(
self,
genome: HypothesisAgentGenome,
system: str,
user: str,
max_tokens: int,
model: str,
) -> CompletionResult:
if self._anthropic is None:
raise RuntimeError("ANTHROPIC_API_KEY required for tier B genomes")
raise RuntimeError(
f"ANTHROPIC_API_KEY required for tier {genome.model_tier.value} genomes"
)
msg = self._anthropic.messages.create(
model=self.model_tier_b,
model=model,
system=system,
messages=[{"role": "user", "content": user}],
temperature=genome.temperature,
@@ -105,6 +145,6 @@ class LLMClient:
text=text,
input_tokens=msg.usage.input_tokens,
output_tokens=msg.usage.output_tokens,
tier=ModelTier.B,
model=self.model_tier_b,
tier=genome.model_tier,
model=model,
)
src/multi_swarm/llm/cost_tracker.py
+4 -1
View File
@@ -8,8 +8,11 @@ from typing import Any
from ..genome.hypothesis import ModelTier
PRICE_PER_M_TOKENS: dict[ModelTier, dict[str, float]] = {
ModelTier.C: {"input": 0.40, "output": 0.40},
ModelTier.S: {"input": 15.00, "output": 75.00},
ModelTier.A: {"input": 3.00, "output": 15.00},
ModelTier.B: {"input": 3.00, "output": 15.00},
ModelTier.C: {"input": 0.40, "output": 0.40},
ModelTier.D: {"input": 0.10, "output": 0.30},
}