Link
Embed
Share
Beginning
This slide
Copy link URL
Copy link URL
Copy iframe embed code
Copy iframe embed code
Copy javascript embed code
Copy javascript embed code
Share
Tweet
Share
Tweet
Slide 1
Slide 1 text
confidencial LLMࢹͷ࠷લઢ IVRy ΤϯδχΞLTେձ 2024/12/11 Moriya Hiroyuki
Slide 2
Slide 2 text
confidencial 2 Զͷ໊લɺߴߍੜ୳ఁ౻৽Ұ
Slide 3
Slide 3 text
confidencial 3 Զͷ໊લɺߴߍੜ୳ఁ౻৽Ұ AIΤϯδχΞͱͯ͠ಇ͖࢝ΊͨԶɺLLMΛͬͯɺͨ͘͞Μ͓ۚΛՔ͍Ͱ͍Δε λʔτΞοϓΛܸͨ͠ɻ
Slide 4
Slide 4 text
confidencial 4 Զͷ໊લɺߴߍੜ୳ఁ౻৽Ұ AIΤϯδχΞͱͯ͠ಇ͖࢝ΊͨԶɺLLMΛͬͯɺͨ͘͞Μ͓ۚΛՔ͍Ͱ͍Δε λʔτΞοϓΛܸͨ͠ɻ ʮͪΐͪΐͬͱ։ൃͨ͠ΒϘϩṶ͚Ͱ͖ΔΜʂʯͱؾ͍ͮͨԶɺىۀͯ͠ɺͻͨ ͢ΒPoCϓϩδΣΫτΛΫϥΠΞϯτʹఏڙ͢Δ͜ͱʹͨ͠ɻ
Slide 5
Slide 5 text
confidencial 5 Զͷ໊લɺߴߍੜ୳ఁ౻৽Ұ AIΤϯδχΞͱͯ͠ಇ͖࢝ΊͨԶɺLLMΛͬͯɺͨ͘͞Μ͓ۚΛՔ͍Ͱ͍Δε λʔτΞοϓΛܸͨ͠ɻ ʮͪΐͪΐͬͱ։ൃͨ͠ΒϘϩṶ͚Ͱ͖ΔΜʂʯͱؾ͍ͮͨԶɺىۀͯ͠ɺͻͨ ͢ΒPoCϓϩδΣΫτΛΫϥΠΞϯτʹఏڙ͢Δ͜ͱʹͨ͠ɻ Զɺഎޙ͔Β͍ۙͮͯ͘Δrate limit੍ݶͱɺLatencyͷѱԽʹؾ͕͍͍ͭͯͳ͔ͬ ͨɻ
Slide 6
Slide 6 text
confidencial 6 Զͷ໊લɺߴߍੜ୳ఁ౻৽Ұ AIΤϯδχΞͱͯ͠ಇ͖࢝ΊͨԶɺLLMΛͬͯɺͨ͘͞Μ͓ۚΛՔ͍Ͱ͍Δε λʔτΞοϓΛܸͨ͠ɻ ʮͪΐͪΐͬͱ։ൃͨ͠ΒϘϩṶ͚Ͱ͖ΔΜʂʯͱؾ͍ͮͨԶɺىۀͯ͠ɺͻͨ ͢ΒPoCϓϩδΣΫτΛΫϥΠΞϯτʹఏڙ͢Δ͜ͱʹͨ͠ɻ Զɺഎޙ͔Β͍ۙͮͯ͘Δrate limit੍ݶͱɺLatencyͷѱԽʹؾ͕͍͍ͭͯͳ͔ͬ ͨɻ ؾ͕͍ͭͨΒɺԶͷϓϩμΫτղ͕૬࣍͗ɺձࣾ࢈ͯ͠͠·͍ͬͯͨ...
Slide 7
Slide 7 text
confidencial 7
Slide 8
Slide 8 text
confidencial 8 ࠓɺ౻৽Ұ܅͕ɺ͜Μͳ݁Λܴ͑ͳ͍ͨΊʹͰ͖Δ͜ͱΛ͓͠͠·͢ɻ
Slide 9
Slide 9 text
confidencial ࣗݾհ 2024/08 ೖࣾ SWEɾػցֶशΤϯδχΞͳͲΛܦݧ LLM͕ίΞʹͳΓͦ͏ͳαʔϏεͩͱࢥͬͯIVRyʹೖࣾ Moriya Hiroyuki 9 AI engineer
Slide 10
Slide 10 text
confidencial IVRyͰͷLLMΛར༻ͨ͠AIର 10 WebsocketΛར༻͠ΤϯυϢʔβʔͱLLM͕ϦΞϧλΠϜʹΓऔΓ͍ͯ͠Δ
Slide 11
Slide 11 text
confidencial LLM Fallback 11 ෳͷLLMΛར༻͢Δ͜ͱΛલఏʹFallbackػߏΛߏங APIͷStatus, Ratelimitσʔλ੍(ཧ੍)ΛͱʹৼΓ͚
Slide 12
Slide 12 text
confidencial LLM Fallback 12 ෳͷLLMΛར༻͢Δ͜ͱΛલఏʹFallbackػߏΛߏங APIͷStatus, Ratelimitσʔλ੍(ཧ੍)ΛͱʹৼΓ͚ ࢹ͢Ε ྑ͍ͷ͡Ό
Slide 13
Slide 13 text
confidencial ํ๏ 1ɿDataDog LLM observability 13 DataDog͕Ӷҙ։ൃதͷLLMࢹʹಛԽͨ͠ػೳɻ Latency, token, promptͳͲΛऔಘͰ͖Δɻ
Slide 14
Slide 14 text
confidencial 14 ʮ͜ΕͰɺOpenAIͷlatency͕ࢹͰ͖ΔΑ͏ʹͳͬͨͥʂʯ
Slide 15
Slide 15 text
confidencial 15 ͋ΕΕʙɺ͓͔͍͠Αʙ
Slide 16
Slide 16 text
confidencial 16 ͋ΕΕʙɺ͓͔͍͠Αʙ ΒͷϓϩμΫτɺfallbackػߏΛ ࣮͍ͯ͠Δͷʹɺ OpenAIͷlatency͔͠ࢹͰ͖ͯͳ͍Αʙ
Slide 17
Slide 17 text
confidencial ํ๏ 2ɿOpenLIT (OpenTelemetry) 17 OpenTelemetryن֨ʹଇͬͨɺLLMࢹʹಛԽͨ͠πʔϧɻ ༷ʑͳLLMΛࢹ͢Δ͜ͱ͕Ͱ͖Δɻ
Slide 18
Slide 18 text
confidencial 18 ʮ͜ΕͰɺ৭ʑͳmodelͷlatency͕ࢹͰ͖ΔΑ͏ʹͳͬͨͥʂʯ
Slide 19
Slide 19 text
confidencial 19 ͋ΕΕʙɺ͓͔͍͠Αʙ
Slide 20
Slide 20 text
confidencial 20 ͋ΕΕʙɺ͓͔͍͠Αʙ Βɺ৭ʑͳϞσϧΛ͏Μ͔ͩΒɺ provider͝ͱʹɺlatencyΛܭଌ͢Δඞཁ͕͋Δͷʹ LiteLLMશମͰͷlatency͔͠औΕͯͳ͍Αʙ
Slide 21
Slide 21 text
confidencial ํ๏ 3ɿDataDog Inferred services 21 DataDogʹࡌ͞ΕͨɺApp֎ͷϦΫΤετΛࢹͯ͘͠ΕΔػߏ
Slide 22
Slide 22 text
confidencial 22 ʮ͜ΕͰɺLiteLLMͰ͍ͬͯΔͯ͢ͷmodelΛࢹͰ͖ΔΑ͏ʹͳͬͨͥʂʯ
Slide 23
Slide 23 text
confidencial 23 ͋ΕΕʙɺ͓͔͍͠Αʙ
Slide 24
Slide 24 text
confidencial 24 ͋ΕΕʙɺ͓͔͍͠Αʙ ΒɺGeminiɺOpenAIͰ̍ͭͷmodelΛ ͏ͱݶΒͳ͍ͷʹɺ ݸผͷmodelͷlatencyΛऔಘ͢Δ͜ͱ Ͱ͖ͯͳ͍Αʙ
Slide 25
Slide 25 text
confidencial ·ͱΊ LLMࢹɺ·ͩ·ͩൃల్্Ͱݟ͕͋Γ·ͤΜʂ AIɾLLMΛ͍͜ͳͯ͠ϓϩμΫτʹೖΕ͍ͯ͘աఔͰɺ ࣗΒ͕Γ։͍͍ͯ͘ඞཁ͕͋Γ·͢ɻ ͥͻҰॹʹAIࢹΛ͍͖ͬͯ·͠ΐ͏ʂ 25