*jUdZddlmZddlZddlZddlZddlZddlZddlm Z ddl m Z m Z ddl mZmZddlmZ ddlmZmZej,eZGd d ZeZdad ed <d$d Zd%dZd&d'dZd(dZ dZ!d)dZ"dZ#d*dZ$dZ%d+dZ&d+dZ'd,dZ(d-dZ) d& d.dZ*d/d0dZ+dZ,d1dZ-dZ.d2dZ/d3d Z0d4d!Z1d"Z2d5d#Z3y#e$re Ze ZYwxYw)6u services/openai_service.py — All OpenAI interactions ===================================================== Single module for both the synchronous "AI compiler / evaluator" surface used by the Coding module AND the asynchronous streaming surface used by the Role Play module (chat, TTS, STT, sentiment, end-of-session report). Key + base URL come from `config.settings` — never hardcoded. ) annotationsN)BytesIO)AnyAsyncGenerator) AsyncOpenAIOpenAI)settings) ChatMessageScenarioceZdZed dZed dZed dZed dZed dZed dZ ed dZ y) _SettingsShimc"tjSN)r OPENAI_API_KEYselfs r/Users/priyanka/Documents/AI_Agent_Assessment/pepole_hub_candidate/people_hub_candidate/services/openai_service.pyopenai_api_keyz_SettingsShim.openai_api_key(s3;3J3J,Jc"tjSr)r OPENAI_BASE_URLrs ropenai_base_urlz_SettingsShim.openai_base_url*s3;3K3K,Krc"tjSr)r OPENAI_MODELrs ropenai_chat_modelz_SettingsShim.openai_chat_model,s3;3H3H,Hrc.tjddS)NOPENAI_TTS_MODELztts-1osgetenvrs ropenai_tts_modelz_SettingsShim.openai_tts_model.3599=OQX3Y,Yrc.tjddS)NOPENAI_TTS_VOICEalloyrrs ropenai_tts_voicez_SettingsShim.openai_tts_voice0r"rc.tjddS)NOPENAI_TTS_FORMATpcmrrs ropenai_tts_formatz_SettingsShim.openai_tts_format2s3599=PRW3X,Xrc.tjddS)NOPENAI_STT_MODELz whisper-1rrs ropenai_stt_modelz_SettingsShim.openai_stt_model4s3599=OQ\3],]rNreturnstr) __name__ __module__ __qualname__propertyrrrr!r&r*r-rrr r 'sd JJ KK HH YY YY XX ]]rr z OpenAI | None _sync_clientclt)ttjtjatS)Napi_keybase_url)r6rr rrr5rr get_clientr;>s,++--  rcTttjtjS)z>Return a fresh AsyncOpenAI client (cheap; no connection held).r8)rr rrr5rr _async_clientr=Hs! '')) rctjjjtj |||ntj |}|jdjjxsdS)Nmodelmessages temperature max_tokensr) r;chat completionscreater rOPENAI_TEMPERATUREchoicesmessagecontent)rArCrBresps r_chatrMPse <   ( ( / /###.#:K@[@[ 0 D <<? " " * * 0b0rctjdd|jddj}t j |S)Nz ```(?:json)?rD```)resubreplacestripjsonloads)rawcleaneds r _extract_jsonrXZs9ff_b#.66ubAGGIG ::g ruYou are a strict compiler and runtime environment. Your ONLY role is to simulate compiler/interpreter output for candidate assessment. ABSOLUTE RULES: 1. DO NOT provide hints, suggestions, fixes, or solutions. 2. DO NOT describe what the code does. 3. DO NOT offer encouragement or commentary. 4. Respond EXACTLY as the real compiler/interpreter would — nothing more. 5. If code has errors: show compiler error messages (file:line:col: error: description). 6. If code runs successfully: output "Program executed successfully." then any stdout. 7. If code is logically wrong but compiles: show runtime output (may be incorrect). Format: raw compiler output only.ctjstjddddSd|d|d|d} t d t d d |d gtj d }d|xsdjv}|xsd|rddSddS#t$r8}tjdt|jdddcYd}~Sd}~wwxYw)uSimulate compilation. Returns compiler-style output. Failure modes are normalised to candidate-safe messages — never echo the upstream OpenAI exception class or HTTP body, which can contain rate-limit URLs, request IDs, or organisation slugs. z.compile_code: OPENAI_API_KEY is not configureduZCode analysis is temporarily unavailable. Your code is saved — please try again shortly.error)outputstatuszCompile and execute this z code for question 'z': ```  ```systemrolerKuserrArCrBzcompile_code failed: %sulThe AI compiler is temporarily unavailable. Your code has not been changed — please try again in a moment.Nzexecuted successfullyrDsuccess) r rlogrZrMCOMPILER_SYSTEMOPENAI_MAX_TOKENS_COMPILE Exceptionwarningtyper1lower)languagecodequestion_titleuser_msgrVexc is_successs r compile_coderrrs  " " BCI   $H:-A.AQRvU   !o>h7 99  $)SYB,=,=,??JiR 9 PP PP   -tCy/A/ABY   s(B C-C;CCaYou are an expert code evaluator for a candidate assessment platform. Evaluate the submitted code objectively. Return ONLY valid JSON, no markdown. JSON structure: { "result": "correct" | "incorrect" | "partial", "score": 0.0 to 1.0, "feedback": "2-3 sentence constructive feedback", "strengths": ["..."], "improvements": ["..."] } Rules: - "correct": code solves the problem fully and efficiently - "partial": code is on the right track but has bugs or edge case failures - "incorrect": code does not solve the problem - Be strict but fair - Never reveal the correct answer - Feedback should be educationally useful without giving away the solutionc d|d|d|d|d } tdtdd|dgtjd }t |}t|jdd xsd }|jdd } t||z} | |jdd|| ||jdgxsg|jdgxsgdS#t $r(}t jd |d d dggd}Yd}~d}~wwxYw)z@Evaluate submitted code. Returns result, score, feedback, marks.z Question: z Language: z Submitted code: ``` r]r^r_rag?rczevaluate_code failed: %s incorrectrbz4Evaluation could not be completed. Please try again.)resultscorefeedback strengths improvementsNrwrvrxrDryrz) ai_result ai_feedbackai_scoreai_marks max_marksryrz) rM EVAL_SYSTEMr OPENAI_MAX_TOKENS_EVALrXrhrerifloatgetround) rlrm question_textrrorVdatarprwrvmarkss r evaluate_coders- ]O$J'j4& 7  !k:h7 66  S!488GS)0S 1E XXh ,F 9u$ %ER0!b17R4:    .4!N   s3C C6C11C6uGenerate a simple visual CAPTCHA challenge for human verification. Return ONLY valid JSON: { "category": "string (e.g. 'vehicles', 'animals', 'food')", "question": "string (e.g. 'Select all images showing vehicles')", "items": [ {"emoji": "🚗", "label": "vehicles", "is_target": true} ] } Rules: - Exactly 9 items in a 3×3 grid - 3–5 items should be the target category - Items should be emojis only - Keep it easy enough for any adult in 5 secondschddddddddddd dddd d ddd dddd dddddddddddddddg dS)Nvehiclesz"Select all images showing vehiclesu🚗T)emojilabel is_targetu🌳natureFu🚕u🍕foodu✈️u🐶animalsu🚂u⛵u🌺)categoryquestionitemsr5r5rr_captcha_fallbackrsq8z Ex Fz Ev F$ Gy Fz Ejt Dx F   rc tdtddddgtjd}t |S#t $r*}t jd|tcYd}~Sd}~wwxYw) zFGenerate a visual emoji CAPTCHA via AI, with a deterministic fallback.r^r_raz:Generate a fresh CAPTCHA challenge with a random category.g?rcu4captcha AI generation failed (%s) — using fallbackN) rMCAPTCHA_SYSTEMr OPENAI_MAX_TOKENS_CAPTCHArXrhreinfor)rVrps rgenerate_captcha_challengersj #!n=.jk 99  S!! # GM ""#s25 A(A#A(#A(cdfd }|dd}d|ddd|dd d |d |d d d|ddd|ddd|dd d|d|ddd|dddS)uBuild a system prompt for a role-play scenario. Accepts either a Scenario model instance OR a plain dict (since the external Role Play API may return arbitrary JSON shapes — see the 'build defensively' direction). All field reads are safe.chtts t|dnj|xs|Sr isinstancedictgetattrrkdefaultscenarios rz&_build_system_prompt.. 0*4Xt*D!T"(,,WX/   r language_nameEnglishz You are ONLY ai_characterz the customerz. You are NOT learner_rolez the candidateza. You must NEVER speak as if you are the candidate or offer solutions as one. SCENARIO CONTEXT: context descriptionz(Practice the conversation realistically.z YOUR CHARACTER: ai_personalityz6Be realistic, emotionally consistent, and challenging.uG CRITICAL RULES — follow these without exception: 1. You are ALWAYS z. The learner is ALWAYS u. 2. NEVER switch roles. NEVER give advice as if you were the candidate. 3. React as your character would — emotionally, realistically, in the moment. 4. LANGUAGE: detect the language the candidate is using in their most recent message and reply in that same language. Match their script (Devanagari/Roman/etc.) and register (formal/casual). If their message is ambiguous, mixed, or empty, default to u. Do not announce or comment on the language switch. 5. Keep responses to 2–3 sentences maximum. Be direct and in-character. 6. Do NOT resolve the situation easily — the learner must earn it through skill. 7. Never say you are an AI. Never mention this is training. 8. Difficulty: difficulty Intermediatez!. 9. Session lasts approximately turnsz exchanges total. rDr5)rg primary_langs` r_build_system_promptrs  A_i0L ..9:;9:; 1]??WXYZhjyXzW{ |#^$L.9:;**+GQ-8K1 Mrc4Kdt|dg}|D]@}t|dddk(rdnd}t|ddxsd}|d vr-|j||dB t}|jj j tj|d d d d{}|23d{}|jdjjxsd} | s3| 97?766y#t$r)} tjd| d| dYd} ~ yd} ~ wwxYww)z/Stream GPT tokens for the AI character's reply.r^r_r`Nra assistantrKrDz[session_start]Tg333333?i,)r@rAstreamrBrCrz&[OpenAI] stream_ai_response failed: %sz'[I'm having trouble responding. Error: ])rrappendr=rErFrG_srrIdeltarKrhrerZ) rAropenai_messagesmr`rKclientrchunkrrps rstream_ai_responserFs$ (5I(5STUO  FD1V;v!Y+1r * * AB ?{{..55&&$ 6  " %MM!$**228bE   6 ? :C@7uA>>>?sgADAC#CC#$C!(C)C!,*C#C#C!!C#"D# D,D DDDcK|xsdj}|sytjxsdj}|xstjxsdj} t }|j jjjtj|||d4d{}|jd23d{}|s tj|jd 47J716dddd{7y#1d{7swYyxYw#t$r!}t j#d ||Yd}~yd}~wwxYww) aStream TTS audio as base64-encoded raw PCM16 chunks (24 kHz mono). Pass ``voice`` to override the env-default OPENAI_TTS_VOICE for a single call. The role-play WebSocket uses this to pair the voice with the chosen avatar's gender (see services/video_engine.py `voice_for_avatar`), so a female avatar never plays over a male voice. When ``voice`` is None we fall back to the env default, which keeps every other caller (chat mode, voice-only mode without avatar selection, dev/test scripts) working unchanged. rDNr)r%?)r@voiceinputresponse_formatspeedi) chunk_sizeasciiz0[OpenAI] stream_tts_chunks failed (voice=%s): %s)rSrr*rkr&r=audiospeechwith_streaming_responserGr! iter_bytesbase64 b64encodedecoderhrerZ)textrfmt chosen_voicerresponserrps rstream_tts_chunksrgs1 JB   D    (5 / / 1C;R00;GBBDL%<<&&>>EE%% F  '22d2C Be **5188AA  BC      % D % %%sA E#AD*3C>4D*7D DDDD(D>D*DD D*DD*ED'D D'#D*&E'D** E3E EEEcK|sy t}t|}d|_ |jjj t j|dd{}t|ddxsd}t|ddxsd}|xsdj}tjd t|||xsd |S7a#t$r\t|}d|_|jjj t j|dd{7}|xsd}d}YwxYw#t$r }tjd |Yd}~yd}~wwxYww) uTranscribe raw audio bytes (webm/opus from MediaRecorder) via Whisper. ``language_code`` is the scenario's primary language. We treat it as a SOFT preference, not a hard filter — if the candidate speaks Hindi during an English-language scenario the transcript should still be in Hindi so the AI character can mirror the language. Implementation: we let Whisper auto-detect (no `language` param) so any language in the candidate's training catalogue works. The scenario language_code is only echoed in the log line for debuggability; it does NOT constrain transcription. This is what enables the "speak any language, AI replies in same language" multilingual flow described in the role-play spec. rDz audio.webm verbose_json)r@filerNrrl?z>[STT] transcribed %d chars detected_lang=%s scenario_lang=%senz$[OpenAI] transcribe_audio failed: %s)r=rnamertranscriptionsrGrr-rrhrSrerlenrZ) raw_bytes language_coder audio_file transcriptr detected_langrps rtranscribe_audiors[ &Y' &  %||::AA)) . B J $J;ArD#J B?F3M !!#  L I}}'<  1 !+J*JO%||::AA)) & B J $"DM $  8#>spED)9CB?&C:D)>E?CAD&D D&#D)%D&&D)) E2E E EEuAnalyse this learner utterance from a professional training roleplay. Return ONLY a valid JSON object — no markdown, no explanation: { "score": , "tone": "", "confidence": "", "structure": "", "empathy": "", "pace": "", "vocabulary": "", "keywords": ["", ""], "tips": [ {"type": "", "label": "", "text": ""} ] } Rules: Always return 2-3 tips. Use exactly one of each type per response. Keep each tip specific to THIS utterance — never generic.c4K|xsdjsiS t}|jjj ddt dd|dgddd {}|j djjxsdj}|jd jd jd j}tj|S7#t$r"}tjd |icYd }~Sd }~wwxYww) z=Analyse a learner utterance and return live coaching metrics.rDz gpt-4o-minir^r_rarir?N```jsonrOz%[OpenAI] analyse_sentiment failed: %s)rSr=rErFrG_SENTIMENT_PROMPTrIrJrK removeprefix removesuffixrTrUrhreri)rrrLrVrps ranalyse_sentimentrs JB    [[,,33!.?@d34  ||A&&..4";;=y)66u=JJ5QWWYzz#   ;SA sHDAC*C(B C*'D(C** D3D D DDDu You are a strict, professional candidate-interview evaluator. Analyse the roleplay conversation and return ONLY a valid JSON object — no markdown, no explanation. CRITICAL SCORING PRINCIPLES — read carefully: • Score the LEARNER's contribution only (the candidate). Never credit the learner for things the AI character said. Never inflate scores out of politeness. Be realistic and evidence-based. • If the learner produced no substantive content (silent, one-word replies, off-topic noise, or fewer than 3 meaningful sentences across the whole session) the overall_score MUST be ≤ 2.0, every dimension score MUST be ≤ 25, and sentiment_timeline values MUST be ≤ 30. • If the learner only partially engaged (some replies but missed the scenario goals, didn't address the AI's concerns, or was vague), cap the overall_score at 5.0 and dimension scores at 55. • Reserve dimension scores ≥ 70 for clear, demonstrated competence in the transcript. Reserve overall_score ≥ 7.5 for strong candidates who actually solved or progressed the scenario. • sentiment_timeline reflects the LEARNER's emotional/communicative quality over time, not the AI's. Silence or disengagement is low sentiment (10–30), not neutral 50. Required schema: { "overall_score": , "verdict": "", "summary": "<2-3 sentence honest paragraph; if the candidate didn't engage, say so plainly>", "communication_style": "", "focus_areas": ["", ""], "dimensions": [ {"name": "", "score": <0-100>, "level": "", "detail": "", "tip": ""} ], "feedback": [ {"type": "", "text": "", "example": "", "priority": ""} ], "key_moments": [ {"turn": , "description": "", "impact": ""} ], "sentiment_timeline": [, ...], "freedom_practice_suggestions": ["", ""] } Dimensions to score (always include all 5): Communication Clarity, Empathy & Tone, Problem-Solving, Professionalism, Active Listening verdict: "pass" if overall_score >= 6.5, else "needs_improvement". freedom_practice_suggestions: 2-3 specific scenarios the learner should rehearse next. c ddddddddddd dd d dd d dddddddddddddg}ddd|ddddg|ddddd d!d"|d#ddd gggd$d%d&gd' S)(a;Realistic report for sessions where the candidate barely engaged. Used as a hard guard so the LLM cannot return a falsely positive report when the transcript shows the learner never spoke. Mirrors the schema in `_REPORT_PROMPT` so downstream code (PDF render, dashboard) sees consistent shape. Communication Claritylowz@No spoken or written responses from the candidate were captured.uPEngage with the scenario — respond to the AI character to demonstrate clarity.rrwleveldetailtipEmpathy & Tonez0Cannot evaluate empathy without candidate input.z9Acknowledge the AI character's concern in your own words.Problem-Solvingrz6No attempt was made to address the scenario's problem.z8Listen for the issue, then propose a concrete next step.Professionalism zCDisengagement during a candidate interview reads as unprofessional.uETreat every prompt as a real interview question — respond promptly.Active ListeningzBNo paraphrasing, clarifying, or follow-up questions were observed.z/Reflect back what the AI said before answering.rneeds_improvementu3The candidate did not engage with the scenario — z. There is no evidence of communication, problem-solving, or interview readiness in this session. A realistic interview outcome would be to not advance the candidate.NonezActive participationzVerbal response under pressureweaknessz-Candidate did not respond during the session.rDhighrjrexamplepriority suggestionzRe-attempt the scenario as the uO and respond to every AI prompt — even short replies are better than silence.) rrrzFCold-open practice: respond within 5 seconds of the AI's first prompt.zFMirror-and-answer drill: paraphrase the AI's question, then answer it.) overall_scoreverdictsummarycommunication_style focus_areas dimensionsrx key_momentssentiment_timelinefreedom_practice_suggestionsr5)rreasonflat_lows r_no_participation_reportr 6s)1Ub d"AEK M#Q%KJ L#R%XW Y$a5WA CH$&A&J= > &.0PQ)X 0!7 ~F[\ 0 1 T T) / rcd}d}g}|D]}t|dd}|dk7rt|ddxsdj}|r"|jdr|jdrZ|j |t j d |Dcgc]}|s| } }|t| z }t| d k\s|d z }|||fScc}w) z>Return (count_of_meaningful_messages, total_words, raw_texts).rr`NrarKrD[rz\s+)rrS startswithendswithrrPsplitr) rAlearner_role_label meaningful total_wordsrVrr`rKwwordss r_learner_engagementrmsJKC q&$' 6> 1i,299;',,S1g6F6Fs6K  7HHVW5;5q5;s5z! u:? !OJ {C '' d?dd@ddAddBdgdCdD|dEddFdGggd.rrr Candidater Characterrrug[Report] no-participation guard fired (meaningful=%d words=%d) — returning realistic low-score reportz0the candidate did not respond during the session)r rKrr`Nraz: rtz(empty session)zCandidate produced z meaningful message(s) and z total word(s). uThis is LOW engagement — apply the strict scoring caps in the system prompt. Do not invent strengths that aren't in the transcript.z Scenario: titlez Role Playz Learner role: z AI character: z Difficulty: rrz Duration: zs z TRANSCRIPT: uv[Report] OpenAI sentiment call → model=%s scenario=%s learner_role=%s duration=%ds meaningful=%d total_words=%dz (no title))rr^r_g333333?ir?rrOug[Report] OpenAI sentiment result ← score=%.1f verdict=%s turns=%d dimensions=%s feedback_count=%drrrrrz(none)rxz,[OpenAI] generate_report_analysis failed: %sg@ru3Report generation failed — fallback values shown.r2midrrrrrrzReport generation failed: z. Please retry.rr)rrrrrxrrr r)rrerr rrjoinrrr=rErFrG_REPORT_PROMPTrIrJrKrSrrrTrUrrrrhrZ)rArduration_secondsrrrrr_rawlinesrrKspeakertranscript_textparticipation_noterrrLrVrdrps ` rgenerate_report_analysisr,s A^[1L^[1L%8,$O!J TQ;?  3   ( E  E !Y+1r * * ")!VT":f"D,, y7),-  ii&;*;O j\)D -' )Q+* T  Qw ,-.%'%'~678%&c  '(  * HH G  '<( / [[,,33&&!n=g64  ||A&&..4";;=y)66u=JJ5QWWYzz#  9 $((?A.3! 4 HHY $ HKHq64(@F(JHK L%)XXl%;%Ar%A C%AqQUU6]%A C Ox $* +  1 (L C   @#F *L02Y[dfg)2Y[dfg*2Y[dfg*2Y[dfg+2Y[dfg &1KC5P_/`F4#%$,.#   soD#L6'AK7K8CK K!K%KK )KL6K K L3AL.(L3)L6.L33L6uXYou are a learning-design expert for a corporate training platform. Given a description, return ONLY a valid JSON object for one roleplay scenario. Required schema (all fields required): { "title": "", "description": "<2-3 sentences>", "context": "<3-5 sentences>", "category": "", "cat_label": "", "learner_role": "", "learner_emoji": "", "ai_character": "", "ai_emoji": "", "ai_personality": "<2-3 sentences>", "difficulty": "", "turns": , "time_limit": "", "scoring": "Guided", "voice_on": true, "language_code": "en", "language_name": "English", "language_flag": "🇬🇧", "featured": false } Return JSON only — no markdown, no explanation.c4K t}|jjjtj dt dd|dgddd{}|jdjjxsd j}|jd jd jd j}tj|S7#t$r*}t j#d |t%d ||d}~wwxYww)zFGenerate a complete scenario definition from a free-text admin prompt.r^r_ragffffff?ir?NrrDrrOz1[OpenAI] generate_scenario_from_prompt failed: %szAI scenario generation failed: )r=rErFrGrr_SCENARIO_GEN_PROMPTrIrJrKrSrrrTrUrhrerZ RuntimeError)promptrrLrVrps rgenerate_scenario_from_promptr1!sM[[,,33&&!.BCf54  ||A&&..4";;=y)66u=JJ5QWWYzz#  M EsKr?s0#  &&4 g!^^ _ # m"1 %'Q\N *'\ 4 $#*%X?F"% "% "%"%R7|?*65p4n(*z B5<MMKHsC CC