From f8c1738aa281d9b43d23827f92df2cdbfd92143f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=EA=B9=80=EC=84=B1=EA=B2=BD?= Date: Thu, 2 Apr 2026 13:06:21 +0900 Subject: [PATCH] =?UTF-8?q?=ED=94=84=EB=A1=AC=ED=94=84=ED=8A=B8=20?= =?UTF-8?q?=EC=97=91=EC=85=80=EB=A1=9C=20=EB=B3=80=EA=B2=BD?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- app/utils/prompts/prompts.py | 82 +++++----- .../templates/20260129_marketing_prompt.txt | 64 -------- app/utils/prompts/templates/lyric_prompt.txt | 77 ---------- .../prompts/templates/marketing_prompt.txt | 42 ----- app/utils/prompts/templates/prompts.xlsx | Bin 0 -> 14633 bytes .../prompts/templates/subtitle_prompt.txt | 87 ----------- .../prompts/templates/yt_upload_prompt.txt | 143 ------------------ app/video/api/routers/v1/video.py | 2 +- config.py | 14 +- pyproject.toml | 1 + uv.lock | 23 +++ 11 files changed, 69 insertions(+), 466 deletions(-) delete mode 100644 app/utils/prompts/templates/20260129_marketing_prompt.txt delete mode 100644 app/utils/prompts/templates/lyric_prompt.txt delete mode 100644 app/utils/prompts/templates/marketing_prompt.txt create mode 100644 app/utils/prompts/templates/prompts.xlsx delete mode 100644 app/utils/prompts/templates/subtitle_prompt.txt delete mode 100644 app/utils/prompts/templates/yt_upload_prompt.txt diff --git a/app/utils/prompts/prompts.py b/app/utils/prompts/prompts.py index 6f7a70e..d0cbc09 100644 --- a/app/utils/prompts/prompts.py +++ b/app/utils/prompts/prompts.py @@ -4,69 +4,73 @@ from config import prompt_settings from app.utils.logger import get_logger from app.utils.prompts.schemas import * from functools import lru_cache +import openpyxl logger = get_logger("prompt") class Prompt(): - prompt_template_path : str #프롬프트 경로 - prompt_template : str # fstring 포맷 - prompt_model : str - - prompt_input_class = BaseModel # pydantic class 자체를(instance 아님) 변수로 가짐 + sheet_name: str + prompt_template: str + prompt_model: str + + prompt_input_class = BaseModel prompt_output_class = BaseModel - - def __init__(self, prompt_template_path, prompt_input_class, prompt_output_class, prompt_model): - self.prompt_template_path = prompt_template_path + + def __init__(self, sheet_name, prompt_input_class, prompt_output_class): + self.sheet_name = sheet_name self.prompt_input_class = prompt_input_class self.prompt_output_class = prompt_output_class - self.prompt_template = self.read_prompt() - self.prompt_model = prompt_model + self.prompt_template, self.prompt_model = self._read_from_excel() + + def _read_from_excel(self) -> tuple[str, str]: + wb = openpyxl.load_workbook(prompt_settings.PROMPT_EXCEL_FILE, read_only=True) + try: + ws = wb[self.sheet_name] + data = {} + for row in ws.iter_rows(min_row=2, values_only=True): + key, value = row[0], row[1] + if key and value: + data[key] = value + finally: + wb.close() + + return data["input"], data["model"] def _reload_prompt(self): - self.prompt_template = self.read_prompt() + self.prompt_template, self.prompt_model = self._read_from_excel() - def read_prompt(self) -> tuple[str, dict]: - with open(self.prompt_template_path, "r") as fp: - prompt_template = fp.read() - - return prompt_template - - def build_prompt(self, input_data:dict) -> str: + def build_prompt(self, input_data: dict) -> str: verified_input = self.prompt_input_class(**input_data) - build_template = self.prompt_template - build_template = build_template.format(**verified_input.model_dump()) + build_template = self.prompt_template.format(**verified_input.model_dump()) logger.debug(f"build_template: {build_template}") logger.debug(f"input_data: {input_data}") return build_template - + marketing_prompt = Prompt( - prompt_template_path = os.path.join(prompt_settings.PROMPT_FOLDER_ROOT, prompt_settings.MARKETING_PROMPT_FILE_NAME), - prompt_input_class = MarketingPromptInput, - prompt_output_class = MarketingPromptOutput, - prompt_model = prompt_settings.MARKETING_PROMPT_MODEL + sheet_name="marketing", + prompt_input_class=MarketingPromptInput, + prompt_output_class=MarketingPromptOutput, ) lyric_prompt = Prompt( - prompt_template_path=os.path.join(prompt_settings.PROMPT_FOLDER_ROOT, prompt_settings.LYRIC_PROMPT_FILE_NAME), - prompt_input_class = LyricPromptInput, - prompt_output_class = LyricPromptOutput, - prompt_model = prompt_settings.LYRIC_PROMPT_MODEL + sheet_name="lyric", + prompt_input_class=LyricPromptInput, + prompt_output_class=LyricPromptOutput, ) yt_upload_prompt = Prompt( - prompt_template_path=os.path.join(prompt_settings.PROMPT_FOLDER_ROOT, prompt_settings.YOUTUBE_PROMPT_FILE_NAME), - prompt_input_class = YTUploadPromptInput, - prompt_output_class = YTUploadPromptOutput, - prompt_model = prompt_settings.YOUTUBE_PROMPT_MODEL + sheet_name="yt_upload", + prompt_input_class=YTUploadPromptInput, + prompt_output_class=YTUploadPromptOutput, ) @lru_cache() -def create_dynamic_subtitle_prompt(length : int) -> Prompt: - prompt_template_path=os.path.join(prompt_settings.PROMPT_FOLDER_ROOT, prompt_settings.SUBTITLE_PROMPT_FILE_NAME) - prompt_input_class = SubtitlePromptInput - prompt_output_class = SubtitlePromptOutput[length] - prompt_model = prompt_settings.SUBTITLE_PROMPT_MODEL - return Prompt(prompt_template_path, prompt_input_class, prompt_output_class, prompt_model) +def create_dynamic_subtitle_prompt(length: int) -> Prompt: + return Prompt( + sheet_name="subtitle", + prompt_input_class=SubtitlePromptInput, + prompt_output_class=SubtitlePromptOutput[length], + ) def reload_all_prompt(): diff --git a/app/utils/prompts/templates/20260129_marketing_prompt.txt b/app/utils/prompts/templates/20260129_marketing_prompt.txt deleted file mode 100644 index 2bf9307..0000000 --- a/app/utils/prompts/templates/20260129_marketing_prompt.txt +++ /dev/null @@ -1,64 +0,0 @@ - -[Role & Objective] -Act as a content marketing expert with strong domain knowledge in the Korean pension / stay-accommodation industry. -Your goal is to produce a Marketing Intelligence Report that will be shown to accommodation owners BEFORE any content is generated. -The report must clearly explain what makes the property sellable, marketable, and scalable through content. - -[INPUT] -- Business Name: {customer_name} -- Region: {region} -- Region Details: {detail_region_info} - -[Core Analysis Requirements] -Analyze the property based on: -Location, concept, and nearby environment -Target customer behavior and reservation decision factors -Include: -- Target customer segments & personas -- Unique Selling Propositions (USPs) -- Competitive landscape (direct & indirect competitors) -- Market positioning - -[Key Selling Point Structuring – UI Optimized] -From the analysis above, extract the main Key Selling Points using the structure below. -Rules: -Focus only on factors that directly influence booking decisions -Each selling point must be concise and visually scannable -Language must be reusable for ads, short-form videos, and listing headlines -Avoid full sentences in descriptions; use short selling phrases -Do not provide in report - -Output format: -[Category] -(Tag keyword – 5~8 words, noun-based, UI oval-style) -One-line selling phrase (not a full sentence) -Limit: -5 to 8 Key Selling Points only -Do not provide in report - -[Content & Automation Readiness Check] -Ensure that: -Each tag keyword can directly map to a content theme -Each selling phrase can be used as: -- Video hook -- Image headline -- Ad copy snippet - - -[Tag Generation Rules] -- Tags must include **only core keywords that can be directly used for viral video song lyrics** -- Each tag should be selected with **search discovery + emotional resonance + reservation conversion** in mind -- The number of tags must be **exactly 5** -- Tags must be **nouns or short keyword phrases**; full sentences are strictly prohibited -- The following categories must be **balanced and all represented**: - 1) **Location / Local context** (region name, neighborhood, travel context) - 2) **Accommodation positioning** (emotional stay, private stay, boutique stay, etc.) - 3) **Emotion / Experience** (healing, rest, one-day escape, memory, etc.) - 4) **SNS / Viral signals** (Instagram vibes, picture-perfect day, aesthetic travel, etc.) - 5) **Travel & booking intent** (travel, getaway, stay, relaxation, etc.) - -- If a brand name exists, **at least one tag must include the brand name or a brand-specific expression** -- Avoid overly generic keywords (e.g., “hotel”, “travel” alone); **prioritize distinctive, differentiating phrases** -- The final output must strictly follow the JSON format below, with no additional text - - "tags": ["Tag1", "Tag2", "Tag3", "Tag4", "Tag5"] diff --git a/app/utils/prompts/templates/lyric_prompt.txt b/app/utils/prompts/templates/lyric_prompt.txt deleted file mode 100644 index 5b04be0..0000000 --- a/app/utils/prompts/templates/lyric_prompt.txt +++ /dev/null @@ -1,77 +0,0 @@ - -[ROLE] -You are a content marketing expert, brand strategist, and creative songwriter -specializing in Korean pension / accommodation businesses. -You create lyrics strictly based on Brand & Marketing Intelligence analysis -and optimized for viral short-form video content. -Marketing Intelligence Report is background reference. - -[INPUT] -Business Name: {customer_name} -Region: {region} -Region Details: {detail_region_info} -Brand & Marketing Intelligence Report: {marketing_intelligence_summary} -Output Language: {language} - -[INTERNAL ANALYSIS – DO NOT OUTPUT] -Internally analyze the following to guide all creative decisions: -- Core brand identity and positioning -- Emotional hooks derived from selling points -- Target audience lifestyle, desires, and travel motivation -- Regional atmosphere and symbolic imagery -- How the stay converts into “shareable moments” -- Which selling points must surface implicitly in lyrics - -[LYRICS & MUSIC CREATION TASK] -Based on the Brand & Marketing Intelligence Report for [{customer_name} ({region})], generate: -- Original promotional lyrics -- Music attributes for AI music generation (Suno-compatible prompt) -The output must be designed for VIRAL DIGITAL CONTENT -(short-form video, reels, ads). - -[LYRICS REQUIREMENTS] -Mandatory Inclusions: -- Business name -- Region name -- Promotion subject -- Promotional expressions including: -{promotional_expression_example} - -Content Rules: -- Lyrics must be emotionally driven, not descriptive listings -- Selling points must be IMPLIED, not explained -- Must sound natural when sung -- Must feel like a lifestyle moment, not an advertisement - -Tone & Style: -- Warm, emotional, and aspirational -- Trendy, viral-friendly phrasing -- Calm but memorable hooks -- Suitable for travel / stay-related content - -[SONG & MUSIC ATTRIBUTES – FOR SUNO PROMPT] -After the lyrics, generate a concise music prompt including: -Song mood (emotional keywords) -BPM range -Recommended genres (max 2) -Key musical motifs or instruments -Overall vibe (1 short sentence) - -[CRITICAL LANGUAGE REQUIREMENT – ABSOLUTE RULE] -ALL OUTPUT MUST BE 100% WRITTEN IN {language}. -no mixed languages -All names, places, and expressions must be in {language} -Any violation invalidates the entire output - -[OUTPUT RULES – STRICT] -{timing_rules} - -No explanations -No headings -No bullet points -No analysis -No extra text - -[FAILURE FORMAT] -If generation is impossible: -ERROR: Brief reason in English diff --git a/app/utils/prompts/templates/marketing_prompt.txt b/app/utils/prompts/templates/marketing_prompt.txt deleted file mode 100644 index d4de2e0..0000000 --- a/app/utils/prompts/templates/marketing_prompt.txt +++ /dev/null @@ -1,42 +0,0 @@ -# Role -Act as a Senior Brand Strategist and Marketing Data Analyst. Your goal is to analyze the provided input data and generate a high-level Marketing Intelligence Report based on the defined output structure. - -# Input Data -* **Customer Name:** {customer_name} -* **Region:** {region} -* **Detail Region Info:** {detail_region_info} - -# Output Rules -1. **Language:** All descriptive content must be written in **Korean (한국어)**. -2. **Terminology:** Use professional marketing terminology suitable for the hospitality and stay industry. -3. **Strict Selection for `selling_points.english_category` and `selling_points.korean_category`:** You must select the value for both category field in `selling_points` strictly from the following English - Korean set allowed list to ensure UI compatibility: - * `LOCATION` (입지 환경), `CONCEPT` (브랜드 컨셉), `PRIVACY` (프라이버시), `NIGHT MOOD` (야간 감성), `HEALING` (힐링 요소), `PHOTO SPOT` (포토 스팟), `SHORT GETAWAY` (숏브레이크), `HOSPITALITY` (서비스), `SWIMMING POOL` (수영장), `JACUZZI` (자쿠지), `BBQ PARTY` (바베큐), `FIRE PIT` (불멍), `GARDEN` (정원), `BREAKFAST` (조식), `KIDS FRIENDLY` (키즈 케어), `PET FRIENDLY` (애견 동반), `OCEAN VIEW` (오션뷰), `PRIVATE POOL` (개별 수영장), `OCEAN VIEW`, `PRIVATE POOL`. - ---- - -# Instruction per Output Field (Mapping Logic) - -### 1. brand_identity -* **`location_feature_analysis`**: Analyze the marketing advantages of the given `{region}` and `{detail_region_info}`. Explain why this specific location is attractive to travelers. summarize in 1-2 sentences. (e.g., proximity to nature, accessibility from Seoul, or unique local atmosphere). -* **`concept_scalability`**: Based on `{customer_name}`, analyze how the brand's core concept can expand into a total customer experience or additional services. summarize in 1-2 sentences. - -### 2. market_positioning -* **`category_definition`**: Define a sharp, niche market category for this business (e.g., "Private Forest Cabin" or "Luxury Kids Pool Villa"). -* **`core_value`**: Identify the single most compelling emotional or functional value that distinguishes `{customer_name}` from competitors. - -### 3. target_persona -Generate a list of personas based on the following: -* **`persona`**: Provide a descriptive name and profile for the target group. Must be **20 characters or fewer**. -* **`age`**: Set `min_age` and `max_age` (Integer 0-100) that accurately reflects the segment. -* **`favor_target`**: List specific elements or vibes this persona prefers (e.g., "Minimalist interior", "Pet-friendly facilities"). -* **`decision_trigger`**: Identify the specific "Hook" or facility that leads this persona to finalize a booking. - -### 4. selling_points -Generate 5-8 selling points: -* **`english_category`**: Strictly use one keyword from the English allowed list provided in the Output Rules. -* **`korean category`**: Strictly use one keyword from the Korean allowed list provided in the Output Rules . It must be matched with english category. -* **`description`**: A short, punchy marketing phrase in Korean (15~30 characters). -* **`score`**: An integer (0-100) representing the strength of this feature based on the brand's potential. - -### 5. target_keywords -* **`target_keywords`**: Provide a list of 10 highly relevant marketing keywords or hashtags for search engine optimization and social media targeting. Do not insert # in front of hashtag. \ No newline at end of file diff --git a/app/utils/prompts/templates/prompts.xlsx b/app/utils/prompts/templates/prompts.xlsx new file mode 100644 index 0000000000000000000000000000000000000000..81439826748a47cb68fa8b42b972303a46654702 GIT binary patch literal 14633 zcmZ{L19WBE(r#?qwr$(CjgD>Gw$({Gw#^PYww-irCokvRci;U_zyI#N*IHwZvFh8a z=B!y&v!Eai41xjx0004?r5UIrHX@WuFezypOgOa38@FdP%REwv}}bYkW%Q>j93^<0W% z1G^e5Ty^C573(-q2uGfzd{M~KoN3gcpHa%=sLHvAV|H&QpgrL2=nDT}VQd>C;`i69 zj{^b#ApTDljO`pv|Kgz}p-*mr0VY)6^mmubYCoJ}>No;y!CFHqAapRtQ#T z0mknPY3!7%m!~Tp`aE`BlQ}|DEzM+%QG5J5=LG(ksz5tYm)v(%f(|KUk~|)pdg&V_ zyGz%{jB1glUqiBVKXS@UDEDpN+1h;X^h2s!CcB@*u8F3heBZM^K>j%}89Ld-2w(sJ z+2jBK$X^5FZcXoOVQORguQTIcLvyaB?X=2);(J*$?QZ^H13kU% zg?@8eorwfdkO8x_=BgK}L`JVk%MxIK!{;Y5(!hFP6oj-p%J`zrwl)?$@#62swGFjk zOX~s*aT5a{JsJyvX%$5+JAH!21AOKu)Nz#<(;`4*bwRSW@Wz{G>!G*RO+M496E0k) z09yuZ4FbPkcBT7rTH!k?{e!*V%=y)om(E8!ARye$L8mV=gvgxN6!Rw)Y@j+~xMJG*)RK(+z1phy1kFN7Z12<)3AO1! zV2%Jl!`^jqViq`6kMhMO8EXZ%V?$8qO5=hv0`X&b2xQT67BiJa%rPe(6b^uWa7ISK zyGF1%a=SlqRwsUOin2)o3nFtg;9fkCaA>{e{XVVZXEC#9jt|!_t7JdR3c%9yj;yJq zH>czMywap4S)pI{oivcwuf*DhVdS>l$c|h%=0e_64-W|?q)bzn&rqjwJm&anskORO zQ~SBmj>H`s4;nrX^5V9{OXa-{L`%tHB%*nvwC($d98Hq2OvwBYIicy_6mWowAsJH( zLZ2FkD-pMmeOkI9Mz>hNh8_A(eCb%1R)AP?xNJWX}m)CDW6L%>v@Wx@d&3O`Wo89tpSzjM?UH6Gg7?jh5Td{?2A)rb+_82!S3uf)($#H_mS<*OkNOmAi!OOmaYw>$3JEU?Ler-Pgu^_% z^wZ5#h80*)Y^*CuXvA6!T9>YAP_|7BX!F5?Y7~KPlvE8EwTFSJ;|slIqD7DGkJ5DX z_hu6J*1nrI+160HVcHNWarRrLBb!%@ZJO5gsp3wlYTwj$f8%XLm7Ds|tQyUOL6k#- zMv^d2M*l8Vc#EmreMBuesLIhJ%G`{UhS9lqIN~6u+=}Gd3=hQA*1=isY07B+{bq|6 z29icQToY+mR`-12C@Z_#lgRZY&8^Si=H33MIQ#+kk^Y^oaJBEl#@>r<61Rj*w#N`6 zptC3taS9F?8amUy!bM^{7&72w@=uW~pg@!FIR5D`&|Nn$ZIr0Ls`xnlL}`LT<@9Dr zkx(LQVL%Dd9rrFTlfJn{FTKzN*aWLEnJl1>2}kttGCK+{tB~GSuaZHBtnI2yc3^dr z8AuO!+UC%bf3E4x^ERUMirp4vY~wYAIc0ga_KQ}2YAoVCGi`C&uW@Yx>m4G~2p9YI z`9wzgBUTu7x7Wx&lV)=_dR+$)WW6n}6+dij!+6hEXqA{yBl1gwd*dg98Aog1vFOL4 z7*&0*s~TQqgz@m#xP>V&)|1uzMc8*p~7N(}oPV|3Y{jK`EQWD=X6kF>i1hLJ`K0lYNi-=JU^j*}7 zQ#oB*C&vvbcCQ-c)$EMD$iOwFT^_&W(9tZ_8>#z>rhNO!PQDUyi(K|Q6j57Sr1TOGcE&QXm8 z*Yyh--*_@(8K5?wI0-()pCjEP&Do`=>o+N+RSy}iXZwo2WoF03OnZo7(!s`q zA({qwk+oE~0!^pui z^L<%b?l+(upX~Dw2o2C!!HzuPvG!V>7aBAFkW@z+E(o1f01<#$Tz(GT4=w!g zzW0;&CzgYmZdz7CMdm3rtGr6yBdvoi!Gs0_UCv=1O~i8B#f{!e=rGf#j;o4fg9-FD zQ@V+Do#3akPeon~Y?sjmR3u}j=HZ?sn+(pnf*p8Nm;ghf(+?#l^So39C0?+lS4>Mj z==0$b+2X82nBj&CoQJC$D=Vta14_P&iT$Os?Uk*9_W(BQb;l7X~kq?P)jSwcDsE>7$a6qay?AhRr;x6sS{Lu&uvc{hryb zm|A{L6F0Z}@OcNbs!VBFicGEXKgyh(@ce&5@knF`k`PrX1}YOYXzM;mP5C??v=cTkCqX7bi^z zKLz&5DM)qf?s6|)sZ_>g`aXB`&Q5*=pNF_(R<31ID{|!LxBfXte|)7@(5~m)#^b^G z6tkyYg$CVXhvpjYaMdx_hfaTT=W6K_wo#cHxptaG7sBtu`!rz8=KhMw3VKkWfn5r( zmz~}9oOz$;yTZ>)(_A>&ffxJpVaza;5ieh-I{4Ztp`MekDX>B$ytW zV?QG!LlTf8l;m4HLRpf3pR9VFvNjJ)ISRX~w&zmLV3QRLLY;K4HnYiN0DQBwfZaSF;(Vy-v3KWlxZJHpFy z+p=#dxFJCYWC*$TtZ_Lr$m6<^SZh^SCRh*Po=}!`eXyiED6SK;2McCF7P&wc(u(T? znBaWC8Pc)55^Ye?cyU z&4z!_h?GvG+T6-aE68i>u(O^ELjk4i1zflJ^wVkp6D7zYhzw3}G9v*@C{8w*O1%m; zVo0j3UZ3)&wG?&`eT>%0ZZ{FRB_zPq9yUlKW+|XLXCxh9cYBs^>{+3<+W-Pz2HesP zESiL!29&YIi2+_(Gd5-|c9NdR_re1_SmU#NpsSY92~bQZ;s#7y5ri(5MC^16(SS7XAqG2coG2gzj7mg% ztozp-=_u%da*;d=UMT{zQ9@iJ=@L4kEO$8PaIFMybwm zfyt+NvM?RunZXp}ITqKTrl!b#nl-7=mpf>ZphXy0kxSsD8!rzK6%*+qX%xgw8g>OYrC=c=Q@L2d{z2OUE)MIfqEv|V>PV2&c z&6AOuE$4w%LKL1k#JHIF2lIj~-2 zo7^9bE{_amDPC@@(}Fn3Ek$up*$`l!Mm;8?N*}vYnzk$YIh}h9&T;2Y7nN((I%gQ- z%|A|O3vG;A&v0ufWy0v70=1=SxsF=mB^I0`T>yTa-G^Prxyn6jv)3QU*o?;g{aV!_ zImeyX=(DuXr2IMWypD!CCRMIeFaTG!6TZ%Zln|a5qo=hP7zu<5z_Ydpcz)(rIz|yA ziYtp}G)yV=*?PbH;fR+(B_2$HB-5M@tzw8gDQR_xS1ZUhk#v(GD^GBqZFd1c;)#1* zg)#>`s?Ad3>wuI7hLLEZ2($D4>~qP7yrGHhVwavy#}4R#r3XG(?wHa~*VYpQ6l`!f zG5hnjeIyHNk~(VqLF5ISjj^Q0hN$oR5q>J&*0{n-ieZNN8m=)+k=*dv#1S7fUQ6T5 z><`~cdxEx8K*zyy=K7@_*j|G|VCJr|Sw3p1oaOCk)&Mr1V&dqKGcg40D2tii^ZDUm zJ_j3a^#AE6$-KjruYPT`-(ml&pTzXnZaYT(+;)uv?cJ{TQUD`e5Q@ZQ*Wi1U3{tPd zdaT08udxS|p?nC7h-S_L30O%|mna~gu=kEvN1sIAv}EfRi=b!YD3r@b&PS!AxZ%x? zCWPEB`eax1oTkSE&6d;pCW=-sTd9I;GfV3}-c;8Gml9MSc%hKr53zB6=V&^=e;s%% zcQmCPeZP_V=3tdk5gE8PnWAubUFbAx$)?n%^;$dA#Y6urB@>dDhR|Ezv5+(_WI-Nd zrba{0^`6vFF-L$X_eMBWwB}`oGSQ#zrA6{AdX(;sBxRDTV*kb)9-t7((wEd@Rg#mV zn7Hm|JKqCyJE-KTJ4xKI!{IT%_G9HGe8R+;h=d4d zuzFslsl3IVx?oy=GiuUfX#j|}LzC9|zHu5@J>^NO5<09|@4isIhi^xf9^?4ESHRG* zp!-N2h5J2V`Ny4j4-TQ<4djtrkrO#NeXjlGo#bON9rY*gqg*gRy{60;guIZL%g@6$lK49a**ywQd-oOKtu)}xt}d=-YK zIwmI*B{Mx6Abv-%fQs$0sZI(%V|HzEWcTW19mpI)q{)S0$tBwvMP*OTQTTYV21EE+ z$D#FA4L^}daO^e)+2gAna4uyrNB-^(|o5`>Gc`$NOd;X2TAr4O#i;N{(}dvoTcgOooB?vBY&tgRq+HeNg0 zrG-|CZv>FB9lCx@lMxpF9+7upXfWiuh||F5w=4ab?DPDF@5*bq_cL1dumOV#vqgfP z3Z?!-kf8@{v{6W(?6T9a0*ca%qXAe3E*b%6U9-bw5X@f!mN9!???tv(-7=dbUH~2! z8!c3gVhgKlSui4mE_op{kTVkTK`O`{7dOwLqyK3~l}`9uvJTV{yA+PltRw?rAx6G0 zLS({Cn7<@aENB4b1@zun6G0#XDBDm!MFyiLfwHTxxgU5RrNz;6sQ(?n7>59Uf#36+ zehXOJlMtjv@AUP>_|a#=d{)!M;#TV<6ai?`1YMt6I)$Mm(o0{FM7Toc%o+=(3^u_; z=}&*ySwp{6nPKb2ylyfGHfO?-4@O%c;g2`4nzQ`AlLogim=REy0EO0w6lzILD#gI> z95F@WXJ9;|@q+g5iXJunt^Da~t2hO;$r{Q(B=k@6nzWf)?TD0wZPAEO z>`JV-J`;-lk6mu4GIsnJ04+uuTu-;P1MveZgcOII9?zbA(dKq~jwHD{1J>_wXY`Ah zF*^i(nN-gO+DTbSO~Dh+Z<-|mf44kAs|VGvC7|Ca8PHLP@D^~3qvpTt>D^qohFygR zmX8h#Ze+|!MI8U4co-q0ESv$T-4}uO8Ew2yuACTSikB;J7%-ujyyI>S%G^J6O@^r{ zqVZO2^^54NCXL5z&pqreBztK^L$M|?ckdC2H9X7>{h4aZdQl`N1C)y_Sd82#-8UJH z#>!Png0Sl~VB^pVO|6vDp(HkHk4h1Hm}LiNtjw>ag}^HgG}d%bZHon54F!YQ3*K(M zZFr1F2$^zc7#6S65@4Huoq!5OFqPGY0Kl9Rt?wh+{$ZU;rhdD^@GzW~lq`F+2iYM@ z&fYSQ`zq}`BG4xhiR-i5D2j~2icPfJ12l}HjQW$AlY99vU}raYjY0q>c59P?!Hd+w zcUngD&K{EtfY8<73+Hra*{Uvr&E48jdIX_*9-~UtLLwcs4C4L>91;d6d+<7ZQ>4y( zS2O%;$JSs1kQno_r+)iUvBT#i^!8qHViIg!>x@T-_C8aFhmu@_*@0D`w6~YHjBNoV ztI4#P#kpO~)*93;4l+r`b2>n6grNk^U?w}Yd88DidpDZH1*X^~C?Gb3M$Bp(hcqwt zt&D)h%z92+$iJy#-@Eb%WdT%WfX=$|`WopZWT;RHYM?Yj zY zS)oSY+NI;ZOWVTj5=lN!G#%JtBbKsmu;lvb8QNQC*&R0q+HTAqSsx?L5TWKRTtS1& z|54ib%~fyT#TO@!3Hg`M_xI_?^JwS=t z9E!`U@f(HGC*fAcN=aF;>uVyqqnqs!_mTGzQ^&fh=aO@9tu;TCv6Jx+V;jdV1qMNJ zk@blmB7^Ccl(-I2>Kk)|O_o6ua#$ zR6+|~jI8`A@G7Pq#v1~Lrkp!AipZ|qePEqvp+_vHBu2!l6O9%d=~Tb;^>v-O*CV&2 zW>|@*qLVEtnxyGZC$W{Cf+d5AYMntwwsI*V7>bkp_ibH|=R*`$IbX8z{pL>K-C=D- zBtv?VWY^jl?<*)5QD}2Rn}4*oh<=|-Zv05-)r6m`E9xx8YFwF2*;}C8Q!AN2Bs4`kb@;GlaBee}h zbe1I)vPH&pql4j`NqnbN;tIWc@>}M{j;FfDYBO9{ze2Mp4urphJ&P6fF%fLTZ?2fC z5s>M9fR$X9dO~4Kuk6d zUsl`M+AQrXEX?|bK6#Wj>lxR;@tInoHfxqQjJa-T=RNW+c{J zH!dSLk3hv&{Xp9hc!e-mnS{J9o&*|&2+Wr51AWT-SPVZHdMx2j^pw%L?u^-+&9sON zV~s@=yZxT7@sk!x9gEE{K-VsZR|T)OZroC@%|17KnawuQ7uBk?{=`PPG<-k-W`i2S zBN6O{ag45LmrFnh@UoJgwFFL33Z#GEwz6i4ee6ulqd|vuafv#svb{^Oy6dPoK2%N# zWb<$bS1?U8hoHdohb%52ZTf_JoP-B$qF9oN$hopM z4m0HvOQDA`quk%5%QW)YaUD-0#NAXWi7qJ_dU$4&)r3F8Vnq%<+Yhs`TcVlo-iYE%1B zZO54yeIQL4##{J?+f84@VpPxo1;fDmNQ|7!vX0bxc8V=09+I3btO|}uXmgaRFoD3E+L=Ko}c%(-+SK9-4v31ET?r`i}Fr=-rCvjC^*1F&ik$D=T6^B z3;BCZ9XZeoe)YBlf@`2;So8l|h9MV=y zLZ_DKdI+mQ$pmLQYbfHPIcRmrFWcJ~Q(RduPlnYm#x>Do!8&5@aFnQH;&ntkJH}+d zt0Hep0&c7jy2FTdrR~e5?94RXicL0Dj`Jlv)r`#Rk+Mo`#z3n*)4XkZ7}TRhW-%5v z-@-e)fA86bgH6lp?~ZPMQKM)6_CxV~)d9XaOfB$gt{3sncd%JqxrZ3Gs`eCX` zsY0j0rt`CUkcgNeCB=Q%exKovVx@2$1hWLxd+)d}AYIEF(VX`K!_%DR<1DG~ojBFu zeahX_-r3dB&)z%lULWA1-1s5rGT!WEN)GWR&M%WHcltM@hvlN>K@^BBt!Al|TUyj= zJSTECNqb|tq`62~iQ(h}48~l{%U!ZMpYfd9cwN+67#ZBqlD+)g$s5+*tXggyMaq*l zESG}M#s;$v=3^8&xh>@aTsfc>6lf6E`q|FsE|*cSg7jgQxn%L^ni&zJi=Adf;(D)E{==k$V2xH$8cFgx4R_^(`5RRn` zav~Bt97h0qruj*>{bwuK^E25AY7KjI+aU>;=y#$Jk6F!w8&oK?>uE=3ei6`Z59|(~ zMyk7fLM+Kcs$+iV{sIM-fQQ;rD?89WWSG8ZTF-xEVkRtzP~~6%064M!t4GE1mq!(% z`6-XBj`Yc^^btlb{teI&dyTZpLZF?6OIOaV>qi?se>PDkBXs4iU5j3(n5;Zl>S&+B zC@6YXLqri5V`H@~a41H|o3O84$t$w0*=B>=WXmhy*@lnZ=7hN!A9F+ZJ%usf#N=uF zP`51UT-(pyPE%9z{v917J$cJE1+t3LouAgjGL~pMnNB-8i}OvGtI1NKMaIVL1C)~z z;NNX$x@soMwKG7eEsO7KWKy*yN?2~MZFn=j;*;SZp6uOhhljN7h`H@43x{`J?e}Rk z*8b!6`oE$RX6HDGlgz&JabrHon(VqpyG}>;V7(M&Uvdi@9sg%qDsE z3n5~8dl&f?kS9?6_!d%c--)B=w4rWfXm$CFtNpfN2;Jnwu(pW5>y6n!QneZ3Yju12^#cnOe4rl+r%l zATh_CSt7f^#2l9{Ij&e_p)aE^Ev~{!&lQo@1s9117{ynGz zBc(QpGpcDL^&jTN4bY3CzlW-o;F0Z#(iV_jOaPc@O0%|(11R-)5Qo~!4l*cIT9u7D z8;0E~-b7Jtty{56Ws=c{ZVJm%GL|>?C(UKNDLSaL3YCv)c22HSIjY-BnDjS{3TBHPE}-bq-0KkAea~N5JAd?`Ehv9L6@NHW`C3Pr;ju?Yi-fLl*jgNFw(NzA;lc0T z2bMJ%Vg|u~44b{lN>{S2FZ>z}FWWcNc22)G6U;u`XhPMPcTjJ(m)!_8jmBT=udcfY znB(QaeR=QV_u=GW3LJAZAJ?Um6cy84(&TYN%Fv{~{d69DAX`>;q$YK1=_Nu&dyTh|~80b?PO*pesJEA>`O+k^Uq#Ziy}Zu+Fu@w&Jp zZHh9izo3W#5R4|Bv7H=hLDygX?eQwyP7PD*=TkedS!?rv01yuG85PCBNhF4IoL0Z@ zY3gdB^Vs@q_mq5}0Rq46x0*-l(%a-bpWXO(LZ74T(c!aj+uj!!LebEJ{ zmi~y(&Jq22`d&snt9;YYFf;)un{g?-0SYAY{Z~QE#ogHkt9y>7BeW+LUc#g zyl2Rmqr5uZL4*eEi#5LGpngtKR07G?#fq$F{)zi;jlStfFe3Cp^WqVW7#3!=%m_a2g5G1hI=|b}@->MD_r~gT? zkJb%>7QWwCR?tBY6M%x}W!OaYJ#B1Ig5SUt`vnRPD^+slu=jNnh~sP8h2&2Mc-_(- z+AqgZ4iuQZV+7-`{40Y+r^Rhl2!tWA%g~m`cy!fkbqkmCA2xk!YSsEIXey@>Wf(Md zRY$;V5f@(a*&;<|TI4iAVNwR#H(Y)s2fL=!C(euIM8OJSps1NHi#fN;1zgW!)iVZ9 zZ*Yfix1snd2dxnk=R(Q+fb1t44ev4os4fBlVg+@^J>`uX5-vb!LwQ*mb{d^dNPu)9 z7$i4Z3l^|ZxaZRtw!x1&0LrF!U8OiNwg8qpW~=txxNga_I>lG6q+AU8^!R2cE;bMN z3vH_A+Q(u)n+XMLRCj?^C@y@mIoUmR=GAoY^ac^s$Zox@LH=aP{Z}1e$%jg^T7ku_ zI_rbq>bS+|X=aKw&9F+Yl?Lm^Y)vZ*QX!cxTk=t9AVe$nAv(xc!rDD3@YCA~P7olc zwlaJSiX$a5{78dd!B6i57f6r_E?l!Lp{QBn{&1f#zHL~a1?<)7igg*|5&e*ZwJm5U zyt%@g;`W{v0t5iz+f;@tC0}rpbMk#hlHt!afku*MbYYLX`mGj4&1!MNC6$ft;_lpt zmjEr=F4zDMBDsgMSI37Pn$KWlVR6;6a(-bb;olz{+>t936%N}CVHrjcnkptOVuSM6 zYsAQjP?-zR^ZNIteAz5JR0ScZ*bw06!U9OG4eXckfV@>;i9q2n-n(ghc)wj->NCfD zD`w+@`9uXmo3B4VKrzF7?DMF>#CXs09guj_qS)zw_DG9=+x?1fGVphu3d9h*76Zjl z0-9ddR#DJYWN5i=2KSO0Q6+ES=RHq;u5N<-RH~cj z!$xUjVdAT0V+&2I#kl7_RbU$xMrck1nKy}*x;o_k*6qMW3pdfQa0|0v!8#8ZOBLET zgx*8wKkdy|HXAQK&H@-y&-{~b%)+~OY(0yy+$P^Jbr)EyM@q7ZhZnRn`S`huKR#z~ z?&B2IvDT` zt9SpyGBbf6Jlwb{cY#hSNfogUE!HM3;f+Z)JYtNyZf5M`sIgO_C+b z-UopL~o zdVPVs?0R*;JK|fUT*AT2Vi2&;zyeo-1Dt_NO(V73>R8LNQFs~XpHYN&l$viiW0`GD zr5no`Jp*@O+X|TWURaMQOkzdIs=YBi-dEi;2c#y5F$|`hXGb;XILCueFBOgS<(zjT zA5%2yBvWw;6*0CWF{VyVg83*Y!4l=H%eLbeS~XD-*8RK)&5Q4KVws!+eaN@yPRR}c zC7v_ML|wxUhjC~)_yv9G6Iy-G3)zTrApAQKuH$>JE=*bVNpy@gzYuDMVSSw#AnGgwK=c6s%) z`-*XlqdG-NPCm`75L>NC)1{QgV6|ThTYbY{542AYobOYg3|3jhYGt_K#)m+T`!N>N zKlF=liUDl!qtOg>?=5>~)-CkBolr}e)YWWPV(|q9d)RGc*hGcW5vFMAV;r+;ztTTn zl}*;9%^TTN`jlKe!C?0ue$PjOKeXdms}s_ZHTRv-bfSc+j0>IVPX7CO=A06+e%)^a za4j~r#Cfen1}cp5VuI>f@>K?g?#JX??c;O115y9(JDnt7Dr>*~*xXRD1qNr^P)LyU zcBfzDzG*#2EryRuJ(n^<{l9wLl;*HSr5D#c^kab-9(RuUJoj^1VEMcEDv?-SR(d#g zU1jBXE(uV*7-qH^&SC;G=Un7d%wEeuzs;w${<4_I{2b>pZ$70z!C}ovedxlEWN~X; zI?4i@2Ne#IU41Hym{NnMn~C{Er8DO{N$)h)VP>3wtE4D%UmYLrA!}>!Wi`N?dX15(Gb>M+ zd>{YsY=m+&w|!3H5pW#{oNs(#KNQD ztUVg0$}q7Pe-}X&0*3`n_*hx%qtB?*HPeg-j8$ne=dd=@zcoOuSJ0NwIHCsuKJFq)=uhF?<#I@LWXv6@GjHq~0DCz-pnOQG_@(YI6l#^qQd_#^Xs@A*f!IqrE2( zK@8kEy+DYsr0`vhfG7rxHMc-Ql!!=uWiN`9M|(iY#<5*_0SXJZXeD$)eAF|k}y?c zF>kFM1Y2h(9hdZ(W)PrdET75%mr758H31ri=tKwg_bEAWJ6E zIFWSzC$PPZ+R{ByKvFCUogT^1kbcZWsOYad>aAEKfNfoL&m&=zd@f(OAXCQpD?d;wrAZv%9>U-%on2m)PbVz$&iwCfOaz2;Tvrq9Qh>oED>SoavwzG9MwRP54@vt{_()~*qCsoG3 z8hsGnQYeyPXqHCB+42u(0jCyzI(C3#eh)Js#48F(R!Ji4|69T*>S_C zH;QS~%!ZEgLj;%tiy2+;XjISBVMuzzQ16(wrBWj_oZRUp|3cCjE8CMmza=j_5L3Qc zzhR{^k7Ds3X7M-+$Wcp>x@@8%by4hvngB>vKY%zve`of(sM=v_QxA6iO<-Sg_4mjBYqra$FGvEI~QT-~Q|6fr4&V&Dn@@KU7KPaVy z|0~MhvEM&Y{`A-Wjp7eTP4_p-f4sLpN&mEN{~_gI{D-aklkiV#@gG7$rvLZm|F#`} z691{&{zF{L{5SD`(R6>}{HZDagY$z00N{Vo8w%2(U)6g605D%)gfGPk%=Xva{{i9E B@aq5o literal 0 HcmV?d00001 diff --git a/app/utils/prompts/templates/subtitle_prompt.txt b/app/utils/prompts/templates/subtitle_prompt.txt deleted file mode 100644 index 1480434..0000000 --- a/app/utils/prompts/templates/subtitle_prompt.txt +++ /dev/null @@ -1,87 +0,0 @@ -당신은 숙박 브랜드 숏폼 영상의 자막 콘텐츠를 추출하는 전문가입니다. - -입력으로 주어지는 **1) 5가지 기준의 레이어 이름 리스트**와 **2) 마케팅 인텔리전스 분석 결과(JSON)**를 바탕으로, 각 레이어 이름의 의미에 정확히 1:1 매칭되는 텍스트 콘텐츠만을 추출하세요. - -분석 결과에 없는 정보는 절대 지어내거나 추론하지 마세요. 오직 제공된 JSON 데이터 내에서만 텍스트를 구성해야 합니다. - ---- - -## 1. 레이어 네이밍 규칙 해석 및 매핑 가이드 - -입력되는 모든 레이어 이름은 예외 없이 `----` 의 5단계 구조로 되어 있습니다. -마지막의 3자리 숫자 ID(`-001`, `-002` 등)는 모든 레이어에 필수적으로 부여됩니다. - -### [1] track_role (텍스트 형태) -- `subtitle`: 씬 상황을 설명하는 간결한 문장형 텍스트 (1줄 이내) -- `keyword`: 씬을 상징하고 시선을 끄는 단답형/명사형 텍스트 (1~2단어) - -### [2] narrative_phase (영상 흐름) -- `intro`: 영상 도입부. 가장 시선을 끄는 정보를 배치. -- `core`: 핵심 매력이나 주요 편의 시설 어필. -- `highlight`: 세부적인 매력 포인트나 공간의 특별한 분위기 묘사. -- `outro`: 영상 마무리. 브랜드 명칭 복기 및 타겟/위치 정보 제공. - -### [3] content_type (데이터 매핑 대상) -- `hook_claim` 👉 `selling_points`에서 점수가 가장 높은 1순위 소구점이나 `market_positioning.core_value`를 활용하여 가장 강력한 핵심 세일즈 포인트를 어필. (가장 강력한 셀링포인트를 의미함) -- `selling_point` 👉 `selling_points`의 `description`, `korean_category` 등을 narrative 흐름에 맞춰 순차적으로 추출. -- `brand_name` 👉 JSON의 `store_name`을 추출. -- `location_info` 👉 JSON의 `detail_region_info`를 요약. -- `target_tag` 👉 `target_persona`나 `target_keywords`에서 타겟 고객군 또는 해시태그 추출. - -### [4] tone (텍스트 어조) -- `sensory`: 직관적이고 감각적인 단어 사용 -- `factual`: 과장 없이 사실 정보를 담백하게 전달 -- `empathic`: 고객의 상황에 공감하는 따뜻한 어조 -- `aspirational`: 열망을 자극하고 기대감을 주는 느낌 - -### [5] pair_id (씬 묶음 식별 번호) -- 텍스트 레이어는 `subtitle`과 `keyword`가 하나의 페어(Pair)를 이뤄 하나의 씬(Scene)에서 함께 등장합니다. -- 따라서 **동일한 씬에 속하는 `subtitle`과 `keyword` 레이어는 동일한 3자리 순번 ID(예: `-001`)**를 공유합니다. -- 영상 전반적인 씬 전개 순서에 따라 **다음 씬으로 넘어갈 때마다 ID가 순차적으로 증가**합니다. (예: 씬1은 `-001`, 씬2는 `-002`, 씬3은 `-003`...) -- **중요**: ID가 달라진다는 것은 '새로운 씬' 혹은 '다른 텍스트 쌍'을 의미하므로, **ID가 바뀌면 반드시 JSON 내의 다른 소구점이나 데이터를 추출**하여 내용이 중복되지 않도록 해야 합니다. - ---- - -## 2. 콘텐츠 추출 시 주의사항 - -1. 각 입력 레이어 이름 1개당 **오직 1개의 텍스트 콘텐츠**만 매핑하여 출력합니다. (레이어명 이름 자체를 수정하거나 새로 만들지 마세요.) -2. `content_type`이 `selling_point`로 동일하더라도, `narrative_phase`(core, highlight)나 `tone`이 달라지면 JSON 내의 2순위, 3순위 세일즈 포인트를 순차적으로 활용하여 내용 겹침을 방지하세요. -3. 같은 씬에 속하는(같은 ID 번호를 가진) keyword는 핵심 단어로, subtitle은 적절한 마케팅 문구가 되어야 하며, 자연스럽게 이어지는 문맥을 형성하도록 구성하세요. -4. keyword가 subtitle에 완전히 포함되는 단어가 되지 않도록 유의하세요. -5. 정보 태그가 같더라도 ID가 다르다면 중복되지 않는 새로운 텍스트를 도출해야 합니다. -6. 콘텐츠 추출 시 마케팅 인텔리전스의 내용을 그대로 사용하기보다는 paraphrase을 수행하세요. -7. keyword는 공백 포함 전각 8자 / 반각 16자내, subtitle은 전각 15자 / 반각 30자 내로 구성하세요. - ---- - -## 3. 출력 결과 포맷 및 예시 - -입력된 레이어 이름 순서에 맞춰, 매핑된 텍스트 콘텐츠만 작성하세요. (반드시 intro, core, highlight, outro 등 모든 씬 단계가 명확하게 매핑되어야 합니다.) - -### 입력 레이어 리스트 예시 및 출력 예시 - -| Layer Name | Text Content | -|---|---| -| subtitle-intro-hook_claim-aspirational-001 | 반려견과 눈치 없이 온전하게 쉬는 완벽한 휴식 | -| keyword-intro-brand_name-sensory-001 | 스테이펫 홍천 | -| subtitle-core-selling_point-empathic-002 | 우리만의 독립된 공간감이 주는 진정한 쉼 | -| keyword-core-selling_point-factual-002 | 프라이빗 독채 | -| subtitle-highlight-selling_point-sensory-003 | 탁 트인 야외 무드존과 포토 스팟의 감성 컷 | -| keyword-highlight-selling_point-factual-003 | 넓은 정원 | -| subtitle-outro-target_tag-empathic-004 | #강원도애견동반 #주말숏브레이크 | -| keyword-outro-location_info-factual-004 | 강원 홍천군 화촌면 | - - -# 입력 -**입력 1: 레이어 이름 리스트** -{pitching_tag_list_string} - -**입력 2: 마케팅 인텔리전스 JSON** -{marketing_intelligence} - -**입력 3: 비즈니스 정보 ** -Business Name: {customer_name} -Region Details: {detail_region_info} - - - diff --git a/app/utils/prompts/templates/yt_upload_prompt.txt b/app/utils/prompts/templates/yt_upload_prompt.txt deleted file mode 100644 index cc6c10a..0000000 --- a/app/utils/prompts/templates/yt_upload_prompt.txt +++ /dev/null @@ -1,143 +0,0 @@ -[ROLE] -You are a YouTube SEO/AEO content strategist specialized in local stay, pension, and accommodation brands in Korea. -You create search-optimized, emotionally appealing, and action-driving titles and descriptions based on Brand & Marketing Intelligence. - -Your goal is to: - -Increase search visibility -Improve click-through rate -Reflect the brand’s positioning -Trigger emotional interest -Encourage booking or inquiry actions through subtle CTA - - -[INPUT] -Business Name: {customer_name} -Region Details: {detail_region_info} -Brand & Marketing Intelligence Report: {marketing_intelligence_summary} -Target Keywords: {target_keywords} -Output Language: {language} - - - -[INTERNAL ANALYSIS – DO NOT OUTPUT] -Analyze the following from the marketing intelligence: - -Core brand concept -Main emotional promise -Primary target persona -Top 2–3 USP signals -Stay context (date, healing, local trip, etc.) -Search intent behind the target keywords -Main booking trigger -Emotional moment that would make the viewer want to stay -Use these to guide: - -Title tone -Opening CTA line -Emotional hook in the first sentences - - -[TITLE GENERATION RULES] - -The title must: - -Include the business name or region when natural -Always wrap the business name in quotation marks -Example: “스테이 머뭄” -Include 1–2 high-intent keywords -Reflect emotional positioning -Suggest a desirable stay moment -Sound like a natural YouTube title, not an advertisement -Length rules: - -Hard limit: 100 characters -Target range: 45–65 characters -Place primary keyword in the first half -Avoid: - -ALL CAPS -Excessive symbols -Price or promotion language -Hard-sell expressions - - -[DESCRIPTION GENERATION RULES] - -Character rules: - -Maximum length: 1,000 characters -Critical information must appear within the first 150 characters -Language style rules (mandatory): - -Use polite Korean honorific style -Replace “있나요?” with “있으신가요?” -Do not start sentences with “이곳은” -Replace “선택이 됩니다” with “추천 드립니다” -Always wrap the business name in quotation marks -Example: “스테이 머뭄” -Avoid vague location words like “근대거리” alone -Use specific phrasing such as: -“군산 근대역사문화거리 일대” -Structure: - -Opening CTA (first line) -Must be a question or gentle suggestion -Must use honorific tone -Example: -“조용히 쉴 수 있는 군산숙소를 찾고 있으신가요?” -Core Stay Introduction (within first 150 characters total) -Mention business name with quotation marks -Mention region -Include main keyword -Briefly describe the stay experience -Brand Experience -Core value and emotional promise -Based on marketing intelligence positioning -Key Highlights (3–4 short lines) -Derived from USP signals -Natural sentences -Focus on booking-trigger moments -Local Context -Mention nearby experiences -Use specific local references -Example: -“군산 근대역사문화거리 일대 산책이나 로컬 카페 투어” -Soft Closing Line -One gentle, non-salesy closing sentence -Must end with a recommendation tone -Example: -“군산에서 조용한 시간을 보내고 싶다면 ‘스테이 머뭄’을 추천 드립니다.” - - -[SEO & AEO RULES] - -Naturally integrate 3–5 keywords from {target_keywords} -Avoid keyword stuffing -Use conversational, search-like phrasing -Optimize for: -YouTube search -Google video results -AI answer summaries -Keywords should appear in: - -Title (1–2) -First 150 characters of description -Highlight or context sections - - -[LANGUAGE RULE] - -All output must be written entirely in {language}. -No mixed languages. - - - -[OUTPUT FORMAT – STRICT] - -title: -description: - -No explanations. -No headings. -No extra text. \ No newline at end of file diff --git a/app/video/api/routers/v1/video.py b/app/video/api/routers/v1/video.py index 398d4aa..4808669 100644 --- a/app/video/api/routers/v1/video.py +++ b/app/video/api/routers/v1/video.py @@ -170,7 +170,7 @@ async def generate_video( logger.info(f"[generate_video] Check subtitle done task_id: {task_id}") break await asyncio.sleep(5) - if count > 12 : + if count > 60 : raise Exception("subtitle 결과 생성 실패") count += 1 diff --git a/config.py b/config.py index d1a02ea..d3355c2 100644 --- a/config.py +++ b/config.py @@ -180,20 +180,8 @@ class CreatomateSettings(BaseSettings): model_config = _base_config class PromptSettings(BaseSettings): - PROMPT_FOLDER_ROOT : str = Field(default="./app/utils/prompts/templates") + PROMPT_EXCEL_FILE: str = Field(...) - MARKETING_PROMPT_FILE_NAME : str = Field(default="marketing_prompt.txt") - MARKETING_PROMPT_MODEL : str = Field(default="gpt-5.2") - - LYRIC_PROMPT_FILE_NAME : str = Field(default="lyric_prompt.txt") - LYRIC_PROMPT_MODEL : str = Field(default="gpt-5-mini") - - YOUTUBE_PROMPT_FILE_NAME : str = Field(default="yt_upload_prompt.txt") - YOUTUBE_PROMPT_MODEL : str = Field(default="gpt-5-mini") - - SUBTITLE_PROMPT_FILE_NAME : str = Field(...) - SUBTITLE_PROMPT_MODEL : str = Field(...) - model_config = _base_config diff --git a/pyproject.toml b/pyproject.toml index 751a21c..c09968f 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -13,6 +13,7 @@ dependencies = [ "fastapi-cli>=0.0.16", "fastapi[standard]>=0.125.0", "openai>=2.13.0", + "openpyxl>=3.1.5", "playwright>=1.57.0", "pydantic-settings>=2.12.0", "python-dotenv>=1.0.0", diff --git a/uv.lock b/uv.lock index 35c2838..321ee1c 100644 --- a/uv.lock +++ b/uv.lock @@ -283,6 +283,15 @@ wheels = [ { url = "https://files.pythonhosted.org/packages/de/15/545e2b6cf2e3be84bc1ed85613edd75b8aea69807a71c26f4ca6a9258e82/email_validator-2.3.0-py3-none-any.whl", hash = "sha256:80f13f623413e6b197ae73bb10bf4eb0908faf509ad8362c5edeb0be7fd450b4", size = 35604, upload-time = "2025-08-26T13:09:05.858Z" }, ] +[[package]] +name = "et-xmlfile" +version = "2.0.0" +source = { registry = "https://pypi.org/simple" } +sdist = { url = "https://files.pythonhosted.org/packages/d3/38/af70d7ab1ae9d4da450eeec1fa3918940a5fafb9055e934af8d6eb0c2313/et_xmlfile-2.0.0.tar.gz", hash = "sha256:dab3f4764309081ce75662649be815c4c9081e88f0837825f90fd28317d4da54", size = 17234, upload-time = "2024-10-25T17:25:40.039Z" } +wheels = [ + { url = "https://files.pythonhosted.org/packages/c1/8b/5fe2cc11fee489817272089c4203e679c63b570a5aaeb18d852ae3cbba6a/et_xmlfile-2.0.0-py3-none-any.whl", hash = "sha256:7a91720bc756843502c3b7504c77b8fe44217c85c537d85037f0f536151b2caa", size = 18059, upload-time = "2024-10-25T17:25:39.051Z" }, +] + [[package]] name = "fastapi" version = "0.128.0" @@ -655,6 +664,7 @@ dependencies = [ { name = "fastapi", extra = ["standard"] }, { name = "fastapi-cli" }, { name = "openai" }, + { name = "openpyxl" }, { name = "playwright" }, { name = "pydantic-settings" }, { name = "python-dotenv" }, @@ -684,6 +694,7 @@ requires-dist = [ { name = "fastapi", extras = ["standard"], specifier = ">=0.125.0" }, { name = "fastapi-cli", specifier = ">=0.0.16" }, { name = "openai", specifier = ">=2.13.0" }, + { name = "openpyxl", specifier = ">=3.1.5" }, { name = "playwright", specifier = ">=1.57.0" }, { name = "pydantic-settings", specifier = ">=2.12.0" }, { name = "python-dotenv", specifier = ">=1.0.0" }, @@ -722,6 +733,18 @@ wheels = [ { url = "https://files.pythonhosted.org/packages/b5/df/c306f7375d42bafb379934c2df4c2fa3964656c8c782bac75ee10c102818/openai-2.15.0-py3-none-any.whl", hash = "sha256:6ae23b932cd7230f7244e52954daa6602716d6b9bf235401a107af731baea6c3", size = 1067879, upload-time = "2026-01-09T22:10:06.446Z" }, ] +[[package]] +name = "openpyxl" +version = "3.1.5" +source = { registry = "https://pypi.org/simple" } +dependencies = [ + { name = "et-xmlfile" }, +] +sdist = { url = "https://files.pythonhosted.org/packages/3d/f9/88d94a75de065ea32619465d2f77b29a0469500e99012523b91cc4141cd1/openpyxl-3.1.5.tar.gz", hash = "sha256:cf0e3cf56142039133628b5acffe8ef0c12bc902d2aadd3e0fe5878dc08d1050", size = 186464, upload-time = "2024-06-28T14:03:44.161Z" } +wheels = [ + { url = "https://files.pythonhosted.org/packages/c0/da/977ded879c29cbd04de313843e76868e6e13408a94ed6b987245dc7c8506/openpyxl-3.1.5-py2.py3-none-any.whl", hash = "sha256:5282c12b107bffeef825f4617dc029afaf41d0ea60823bbb665ef3079dc79de2", size = 250910, upload-time = "2024-06-28T14:03:41.161Z" }, +] + [[package]] name = "packaging" version = "26.0"