Reshare — abun ciki daga Julien Mutzenhardt
Tarar €17,000 saboda JSON da ba a tsara shi da kyau ba: kuskuren da ya yi sanadiyyar mutuwa tare da LLMs!
Wani mai haɓaka software ya haifar da babban jinkiri wajen sarrafa muhimman fayiloli na kamfaninsa. Dalili? Kuskuren haɗa kai mai sauƙi. 
Duk da haka ya bayyana a fili a cikin umarninsa: "Amsa da JSON." Matsalar ita ce, samfurin, cikin ladabi sosai, ya amsa: "Tabbas, ga bayanin:", kafin ya saka lambar sa.
Sakamakon: mai binciken, wanda yake tsammanin abin JSON kai tsaye, ya faɗi cikin madauki. 
Darasi: Umarni umarni ne kawai, ba wajibi ba.
Kamar ka gaya wa karenka kada ya taɓa kaza a kan tebur, sannan ka kwanta barci. Babu wani abu da ke tilasta wa samfurin basirar wucin gadi yin biyayya sosai ga tsarin da aka nema.
Maganin gaske: "Ƙuntataccen Ƙaddamarwa"
Maimakon fatan samfurin ya bi umarnin rubutunka, ba shi "JSON Schema" (ainihin tsarin amsar da ake tsammani, don bayyana misali cewa adadi dole ya zama lamba).
Yaya yake aiki a ɓoye?
Maimakon barin LLM ya zaɓi kalma ta gaba daga dukan ƙamus ɗinsa, injin ƙirƙira yana sake ƙididdige "tokens" da aka yarda da su a ainihin lokaci. Misali, don fara abin JSON, zai ba da izini kawai buɗewar brace {. Samfurin ba shi da wani zaɓi. Ta haka tsarin ya zama mai inganci ta hanyar gini, wanda ke guje wa ƙaddamar da dubban sabbin yunƙurin da ba dole ba (sake gwadawa)! 
Yaya za a aiwatar da shi ba tare da sake rubuta wani dodo mai rikitarwa ba?
Kawai aika JSON Schema ɗinka zuwa APIs na yau da kullun (kamar OpenAI ko Anthropic), waɗanda ke kula da sarrafa shi. Kuma idan kana karɓar bakuncin samfurin AI naka a gida, injuna kamar vLLM suna kula da shi cikin sauƙi.
Daina yin duk abin da ya faru a cikin samarwa kuma ka tabbatar da bayanan fitarwa daga LLMs ɗinka! 
Takaitaccen wannan labarin an ƙirƙira shi da AI.
Kai kuma, ka taɓa ganin tsarin ya fashe a tsakiyar tafiya saboda amsar da ba a zata ba daga basirar wucin gadi?
#ArtificialIntelligence #Development #JSON #LLM #OpenAI #VLLM #Programming #Tech
Idan kana son tattauna batun da ni ko yin amfani da ƙwarewata, za ka iya tuntuɓar ni ta PM ko kiran bidiyo.
Majiyoyin LinkedIn
- Bayanan martaba: Pierre Jean Duvivier dit Sage (LinkedIn)
- Marubucin da aka raba: Julien Mutzenhardt (LinkedIn)
Rubutun 🚨 Tarar €17,000 saboda JSON da ba a tsara shi da kyau ba: kuskuren da ya yi sanadiyyar mutuwa tare da LLMs! 🚨 ya fara bayyana a les plumes qui mordent.
Tambayoyin da ake yawan yi
FAQs 17000 in Penalties for a Badly Formatted JSON The Fatal Mistake with LLMs
1 What is this 17000 penalty about
It refers to a realworld case where an AI system produced malformed JSON output which broke a downstream process and led to a financial penalty of 17000 Its used as a cautionary example of why output formatting matters
2 What is JSON
JSON is a simple text format for structuring data as keyvalue pairs lists and nested objects Its the standard way software exchanges data
3 What does badly formatted JSON mean
JSON that breaks the syntax rules for example missing quotes trailing commas unescaped characters unbalanced brackets or extra text beforeafter the object Parsers reject it
4 Why do LLMs produce invalid JSON
LLMs generate text probabilistically They may add explanations use markdown code fences hallucinate fields truncate output or slightly deviate from the schema especially without strict instructions or validation
5 How can a formatting error cost 17000
If a system blindly trusts LLM output invalid JSON can crash automated pipelines cause wrong transactions trigger contractSLA penalties or force manual fixes One small syntax error can cascade into real financial loss
6 Whats the most common JSON mistake LLMs make
Wrapping the JSON in markdown or adding conversational text around it Other frequent issues trailing commas single quotes and truncated output
7 How do I get an LLM to return valid JSON
Ask explicitly for JSON only no explanation provide a schema or example use low temperature and most importantly validate and retry on failure
8 What is JSON schema validation
A way to check that JSON matches an expected structure Tools like JSON Schema validators catch errors before they reach production
9 Should I trust LLM output without checking it
No Always validate LLM output programmatically Treat it as untrusted input just like usersubmitted data
10 What are structured outputs or JSON mode in LLMs
Features offered by some model providers that constrain
