Reġistrar u Osservabbiltà
F'din il-paġna
Reġistrar u Osservabbiltà
Fil-produzzjoni, trid tkun taf eżattament x’inhu jiġri bis-sejħiet tal-API tiegħek: il-latency, it-tokens użati, l-iżbalji u l-ispejjeż. Ir-reġistrar strutturat u l-osservabbiltà jagħtuk dik il-viżibbiltà. Id-differenza ssir ċara mill-ewwel mument li utent jirrapporta, “L-assistent kien bil-mod minn dan il-għodu.” Mingħajr strumentazzjoni, tkun imwaħħal tgħodd bl-għajn. Biha, tkun tista’ tara fi tletin sekonda li l-latency p95 irdoppjat wara bidla fil-prompt li ttripplikat id-daqs tal-kuntest.
Reġistrar Bażiku Strutturat
Agħmilha drawwa li tkejjel madwar is-sejħa, mhux ġo fiha. Irreġistra l-arloġġ qabel, aqrah mill-ġdid wara, u emetti linja waħda li tiddeskrivi l-avveniment kollu. JSON mhux biss għad-dehra: jħalli l-aggregatur tiegħek jiffiltra fuq status, isegwi duration_ms maż-żmien, jew ifittex response_id speċifiku—xi ħaġa li test liberu qatt ma tista’ tagħmel. Il-blokk except jipproduċi l-istess struttura eżatta, inkluż it-tip ta’ eċċezzjoni u l-messaġġ, imbagħad jerġa’ jqajjemha. Ir-reġistrar mhuwiex trattament.
import logging
import json
import time
from openai import OpenAI
logging.basicConfig(
level=logging.INFO,
format="%(asctime)s %(levelname)s %(message)s"
)
logger = logging.getLogger("openai_api")
client = OpenAI()
def appel_avec_logging(prompt: str, model: str = "gpt-5.6-terra") -> str:
"""Sejħa tal-API b'reġistrar strutturat sħiħ."""
start = time.monotonic()
request_id = None
try:
response = client.responses.create(
model=model,
input=prompt
)
duree = time.monotonic() - start
logger.info(json.dumps({
"event": "api_call",
"status": "success",
"model": response.model,
"response_id": response.id,
"input_tokens": response.usage.input_tokens,
"output_tokens": response.usage.output_tokens,
"total_tokens": response.usage.total_tokens,
"duration_ms": round(duree * 1000),
"prompt_length": len(prompt),
}))
return response.output_text
except Exception as e:
duree = time.monotonic() - start
logger.error(json.dumps({
"event": "api_call",
"status": "error",
"error_type": type(e).__name__,
"error_message": str(e),
"model": model,
"duration_ms": round(duree * 1000),
"prompt_length": len(prompt),
}))
raise
resultat = appel_avec_logging("Hello!")
Innota li nirreġistraw prompt_length, mhux il-prompt innifsu. Dik hija għażla tad-disinn, mhux żball: it-tul huwa biżżejjed biex tikkorrelata mal-latency, mingħajr ma tesporta data tal-utent għas-sistema tar-reġistrar tiegħek.
Middleware tar-Reġistrar
Li tirrepeti dan il-blokk f’kull funzjoni li ssejjaħ l-API jiggarantixxi li se tinsa xi mkien. Aħjar li ddawwar il-klient darba għal kollox. Il-klassi tad-data APIMetrics tiddefinixxi l-kuntratt ta’ x’inkejjlu, u ObservableClient jimla dik ir-rekord u jarkivjah, kemm jekk is-sejħa tirnexxi jew le. Il-metodu rapport imbagħad ibiddel l-istorja f’metriki direttament azzjonabbli—rata ta’ suċċess, latency medja, latency p95.
Dak in-numru tal-aħħar huwa l-aktar importanti. Medja ta’ 450 ms tista’ taħbi l-wieħed minn għoxrin utent li qed jistenna tliet sekondi; p95 jiżvelaha. Billi tissortja d-durati u tieħu l-valur fil-perċentil 95, tiddeskrivi l-esperjenza reali tal-aktar utenti attivi li qed jiġu servuti l-agħar.
from dataclasses import dataclass, field
from typing import Optional
import time
import json
@dataclass
class APIMetrics:
"""Metriki miġbura għal kull sejħa."""
model: str = ""
input_tokens: int = 0
output_tokens: int = 0
total_tokens: int = 0
duration_ms: int = 0
status: str = "unknown"
error: Optional[str] = None
response_id: Optional[str] = None
class ObservableClient:
"""Klient OpenAI b'osservabbiltà integrata."""
def __init__(self):
self.client = OpenAI()
self.metrics_history: list[APIMetrics] = []
def create(self, model: str, input: str, **kwargs) -> object:
start = time.monotonic()
metrics = APIMetrics(model=model)
try:
response = self.client.responses.create(
model=model,
input=input,
**kwargs
)
metrics.status = "success"
metrics.response_id = response.id
metrics.input_tokens = response.usage.input_tokens
metrics.output_tokens = response.usage.output_tokens
metrics.total_tokens = response.usage.total_tokens
metrics.duration_ms = round((time.monotonic() - start) * 1000)
self.metrics_history.append(metrics)
return response
except Exception as e:
metrics.status = "error"
metrics.error = str(e)
metrics.duration_ms = round((time.monotonic() - start) * 1000)
self.metrics_history.append(metrics)
raise
def rapport(self) -> dict:
"""Iġġenera rapport tal-metriki miġbura."""
if not self.metrics_history:
return {"message": "No calls recorded"}
total_calls = len(self.metrics_history)
erreurs = sum(1 for m in self.metrics_history if m.status == "error")
total_tokens = sum(m.total_tokens for m in self.metrics_history)
durees = [m.duration_ms for m in self.metrics_history]
return {
"total_appels": total_calls,
"erreurs": erreurs,
"taux_succes": f"{((total_calls - erreurs) / total_calls) * 100:.1f}%",
"total_tokens": total_tokens,
"latence_moyenne_ms": round(sum(durees) / len(durees)),
"latence_p95_ms": sorted(durees)[int(len(durees) * 0.95)],
}
# Użu
obs_client = ObservableClient()
obs_client.create("gpt-5.6-terra", "Hello!")
obs_client.create("gpt-5.6-terra", "How are you?")
obs_client.create("gpt-5.6-terra", "Thanks!")
print(json.dumps(obs_client.rapport(), indent=2))
# Riżultat:
# {
# "total_appels": 3,
# "erreurs": 0,
# "taux_succes": "100.0%",
# "total_tokens": 95,
# "latence_moyenne_ms": 450,
# "latence_p95_ms": 520
# }
Traċċar tas-Sejħiet bil-Kuntest
F’arkitettura distribwita, sejħa tal-API qatt ma teżisti waħedha: tiġi attivat minn talba HTTP, li hi nnifisha ġejja minn azzjoni tal-utent, u xi kultant segwita minn żewġ sejħiet oħra. It-trace_id huwa l-fil li jorbot dawn l-avvenimenti flimkien. Billi tippagah mill-layer tal-web tiegħek—hawnhekk user-123-req-456—tkun tista’ tirrikostruwixxi sessjoni sħiħa fir-reġistri tiegħek minn lment wieħed tal-klijent. Jiġu emessi żewġ linji, waħda fil-bidu u waħda fit-tmiem, li tgħin ukoll biex tindividwa sejħiet li qatt ma rritornaw.
import uuid
class TracingClient:
"""Klient b'traċċar biex jikkorrelata t-talbiet."""
def __init__(self):
self.client = OpenAI()
def create_with_trace(self, prompt: str, trace_id: str = None,
**kwargs) -> tuple:
"""Sejħa b'trace_id għall-korrelazzjoni."""
if trace_id is None:
trace_id = str(uuid.uuid4())[:8]
start = time.monotonic()
logger.info(json.dumps({
"event": "api_request_start",
"trace_id": trace_id,
"prompt_preview": prompt[:100],
}))
response = self.client.responses.create(
model=kwargs.get("model", "gpt-5.6-terra"),
input=prompt,
**{k: v for k, v in kwargs.items() if k != "model"}
)
duration = time.monotonic() - start
logger.info(json.dumps({
"event": "api_request_complete",
"trace_id": trace_id,
"response_id": response.id,
"tokens": response.usage.total_tokens,
"duration_ms": round(duration * 1000),
}))
return response, trace_id
# Użu b'trace_id biex tikkorrelata fir-reġistri
tracing = TracingClient()
response, trace = tracing.create_with_trace(
"Summarize logging best practices.",
trace_id="user-123-req-456"
)
print(f"Trace: {trace} | Response: {response.output_text[:80]}...")
Oqgħod attent ma’ prompt_preview: dawk l-ewwel mitt karattru jmorru dritt fir-reġistri tiegħek hekk kif inhuma. Aċċettabbli fl-iżvilupp, neħħih malli tibda tgħaddi data reali.
Esportazzjoni għal Sistema ta’ Monitoraġġ
Storja fil-memorja tisparixxi meta l-proċess jerġa’ jibda. Biex timmonitorja servizz maż-żmien, għandek tesponi l-metriki għal għodda esterna. Il-mudell universali jiddependi fuq żewġ familji: counters, li jiżdiedu biss u jiġu derivati f’rati, u histograms, li jżommu distribuzzjoni biex joħorġu l-perċentili. Din il-klassi hija intenzjonalment newtrali—għandek biss tgħaqqad get_metrics ma’ endpoint biex Prometheus jiskrapjah, jew timbotta d-dizzjunarju lejn Datadog.
# Mudell għall-esportazzjoni lejn Prometheus, Datadog, eċċ.
class MetricsExporter:
"""Esportatur tal-metriki kompatibbli mas-sistemi ta' monitoraġġ."""
def __init__(self):
self.counters = {
"api_calls_total": 0,
"api_errors_total": 0,
"tokens_consumed_total": 0,
}
self.histograms = {
"api_latency_ms": [],
}
def record_success(self, tokens: int, latency_ms: int):
self.counters["api_calls_total"] += 1
self.counters["tokens_consumed_total"] += tokens
self.histograms["api_latency_ms"].append(latency_ms)
def record_error(self):
self.counters["api_calls_total"] += 1
self.counters["api_errors_total"] += 1
def get_metrics(self) -> dict:
latencies = self.histograms["api_latency_ms"]
return {
**self.counters,
"latency_avg_ms": round(sum(latencies) / max(len(latencies), 1)),
"error_rate": self.counters["api_errors_total"] /
max(self.counters["api_calls_total"], 1),
}
exporter = MetricsExporter()
# Sejjaħ exporter.record_success() jew exporter.record_error()
# wara kull sejħa tal-API
Punti Ewlenin
- Irreġistra kull sejħa b’: mudell, tokens, latency, status, response_id
- Uża JSON strutturat għar-reġistri li l-għodod tiegħek jistgħu janalizzaw
- Iġbor metriki ewlenin: rata ta’ suċċess, latency p95, konsum tat-tokens
- Żid
trace_idbiex tikkorrelata t-talbiet f’sistemi distribwiti - Esporta l-metriki għas-sistema ta’ monitoraġġ tiegħek (Prometheus, Datadog, eċċ.)
- Qatt tirreġistra l-kontenut tal-prompt fil-produzzjoni (data sensittiva)