Anton_wireframe/app/services/__pycache__/querying.cpython-312.pyc

Ë
Ç"
iö0ãóà—dd
lZdd
l
Z
dd
lZdd
lZddlmZmZ
ddlmZ
ddl	m
Z
mZmZ
ddl
mZ
ddlmZ
ddlmZmZmZmZ
ddlmZ
dd	lmZ
dd
lmZ
ej8e«
ZGd„d«Zy
)
éN)ÚListÚOptional)
Úget_db)Ú	FundTableÚ
InvestorTableÚProjectTable)
ÚChatPromptTemplate)
Ú
ChatOpenAI)ÚCompanyMinimalÚInvestmentResponseÚPaginatedResponseÚ
SectorMinimal)
Útext)
Úselectinload)
Ú(calculate_project_investor_compatibilitycó”—eZ
dZd
„Zdedefd„Z	ddedeedee	fd„Z
	ddedeedee	fd„Z	dd	eedeedee	fd
„Z
y)ÚQueryProcessorc
ó˜—t
tjd
«
ddd¬«|_i|_tjddg«
|_y)NÚOPENROUTER_API_KEYzhttps://openrouter.ai/api/v1zopenai/gpt-4o-minir)Úapi_keyÚbase_urlÚmodelÚtemperature)Úsystemu%You are a SQL expert. Generate a SQLite query to find fund IDs based on user requirements.

Database Schema:
- funds: id, fund_name, investor_id, check_size_lower, check_size_upper, geographic_focus
- fund_sectors: fund_id, sector_id
- fund_investment_stages: fund_id, stage_id
- sectors: id, name
- investment_stages: id, name
- investors: id, name, aum

IMPORTANT RULES:
1. ALWAYS return ONLY fund IDs (funds.id) - use SELECT DISTINCT f.id
2. For geography: Be FLEXIBLE - use OR with variations and partial matches
   - 'Europe' → WHERE geographic_focus LIKE '%Europe%' OR geographic_focus LIKE '%European%'
   - 'America' → WHERE geographic_focus LIKE '%America%' OR geographic_focus LIKE '%US%' OR geographic_focus LIKE '%United States%'
   - 'Asia' → WHERE geographic_focus LIKE '%Asia%' OR geographic_focus LIKE '%Asian%'
   - If no geography specified, DON'T filter by geography
3. For stages: Use LEFT JOIN and LIKE for flexible matching with synonyms
   - 'Seed' → s.name LIKE '%Seed%' OR s.name LIKE '%Pre-Seed%' OR s.name LIKE '%Early%'
   - 'Series A' → s.name LIKE '%Series A%' OR s.name LIKE '%A%'
   - 'Growth' → s.name LIKE '%Growth%' OR s.name LIKE '%Late%' OR s.name LIKE '%Expansion%'
   - If stage not specified, include ALL funds
4. For sectors: Use LEFT JOIN and include related terms with OR
   - 'Fintech' → sec.name LIKE '%Fintech%' OR sec.name LIKE '%Finance%' OR sec.name LIKE '%Financial%'
   - 'AI' → sec.name LIKE '%AI%' OR sec.name LIKE '%Artificial Intelligence%' OR sec.name LIKE '%Machine Learning%' OR sec.name LIKE '%ML%'
   - 'Healthcare' → sec.name LIKE '%Healthcare%' OR sec.name LIKE '%Health%' OR sec.name LIKE '%Medical%' OR sec.name LIKE '%Biotech%'
5. For check size filters (be flexible with ranges):
   - "under X" → WHERE (check_size_upper <= X OR check_size_upper IS NULL)
   - "over X" → WHERE (check_size_lower >= X OR check_size_lower IS NULL)
   - "between X and Y" → WHERE check_size_lower >= X AND check_size_upper <= Y
6. Use LEFT JOIN for stages and sectors so funds without tags still match
7. Use DISTINCT to avoid duplicates from joins
8. Be INCLUSIVE - use OR conditions to cast a wider net
9. If query is very simple (e.g., just "seed stage"), don't add unnecessary filters
10. Return a single, complete SELECT query

Example Queries:
Q: "Seed stage investors in Europe"
A: SELECT DISTINCT f.id FROM funds f 
   LEFT JOIN fund_investment_stages fis ON f.id = fis.fund_id 
   LEFT JOIN investment_stages s ON fis.stage_id = s.id 
   WHERE (s.name LIKE '%Seed%' OR s.name LIKE '%Pre-Seed%' OR s.name LIKE '%Early%' OR s.id IS NULL)
   AND (f.geographic_focus LIKE '%Europe%' OR f.geographic_focus LIKE '%European%')

Q: "Fintech investors with check size under 5 million"
A: SELECT DISTINCT f.id FROM funds f 
   LEFT JOIN fund_sectors fs ON f.id = fs.fund_id 
   LEFT JOIN sectors sec ON fs.sector_id = sec.id 
   WHERE (sec.name LIKE '%Fintech%' OR sec.name LIKE '%Finance%' OR sec.name LIKE '%Financial%' OR sec.id IS NULL)
   AND (f.check_size_upper <= 5000000 OR f.check_size_upper IS NULL)

Q: "Seed stage investors"
A: SELECT DISTINCT f.id FROM funds f 
   LEFT JOIN fund_investment_stages fis ON f.id = fis.fund_id 
   LEFT JOIN investment_stages s ON fis.stage_id = s.id 
   WHERE s.name LIKE '%Seed%' OR s.name LIKE '%Pre-Seed%' OR s.name LIKE '%Early%'

Q: "Growth stage investors"
A: SELECT DISTINCT f.id FROM funds f 
   LEFT JOIN fund_investment_stages fis ON f.id = fis.fund_id 
   LEFT JOIN investment_stages s ON fis.stage_id = s.id 
   WHERE s.name LIKE '%Growth%' OR s.name LIKE '%Late%' OR s.name LIKE '%Expansion%' OR s.name LIKE '%Series C%' OR s.name LIKE '%Series D%'

Q: "AI investors in America"
A: SELECT DISTINCT f.id FROM funds f 
   LEFT JOIN fund_sectors fs ON f.id = fs.fund_id 
   LEFT JOIN sectors sec ON fs.sector_id = sec.id 
   WHERE (sec.name LIKE '%AI%' OR sec.name LIKE '%Artificial Intelligence%' OR sec.name LIKE '%Machine Learning%' OR sec.name LIKE '%ML%')
   AND (f.geographic_focus LIKE '%America%' OR f.geographic_focus LIKE '%US%' OR f.geographic_focus LIKE '%United States%' OR f.geographic_focus LIKE '%USA%')

Q: "Healthcare investors"
A: SELECT DISTINCT f.id FROM funds f 
   LEFT JOIN fund_sectors fs ON f.id = fs.fund_id 
   LEFT JOIN sectors sec ON fs.sector_id = sec.id 
   WHERE sec.name LIKE '%Healthcare%' OR sec.name LIKE '%Health%' OR sec.name LIKE '%Medical%' OR sec.name LIKE '%Biotech%' OR sec.name LIKE '%Pharma%'

IMPORTANT: Use LEFT JOIN so funds without sector/stage tags can still match. Include synonym terms with OR for better recall.

Return ONLY the SQL query, no explanations or markdown.)Úuserz
{question})r
ÚosÚgetenvÚllmÚquery_cacher	Ú
from_messagesÚ
sql_prompt)
Úselfs
 úL/home/oluwasanmi/Documents/Work/MKD/anton_wireframe/app/services/querying.pyÚ__init__zQueryProcessor.__init__sV€ÜÜ—I‘IÐ2Ó3Ø3Ø&Øô	
ˆŒðˆÔô-×:Ñ:ðQ
ðd'ðgT

óV

ˆ�óÚquestionÚreturncóœ—t
j|
j«j«j	««
j«S)
z,Generate cache key from normalized question.)ÚhashlibÚmd5ÚlowerÚstripÚencodeÚ	hexdigest)r"r&s  r#Ú_get_cache_keyzQueryProcessor._get_cache_key~s3€ä�{‰{˜8Ÿ>™>Ó+×1Ñ1Ó3×:Ñ:Ó<Ó=×GÑGÓIÐIr%NÚ
project_idcƒó`K
—t
j|j|
|«ƒd
{–—†S7Œ­w
)zrAsync wrapper for process_query. Runs blocking work in a thread to avoid
        blocking the event loop.
        N)ÚasyncioÚ	to_threadÚ_process_query_sync)r"r&r0s   r#Ú
process_queryzQueryProcessor.process_query‚s)èø€ô×&Ñ& t×'?Ñ'?ÀÈ:ÓV×VÐVÐVús‚%.
§
,¨.
c	ó2—|j
|
«
}||jvr(|j|}tjd
|›�«

n¨|jj|
¬«
}|jj|«
}|jj«}|jdd«jdd«j«}||j|<tjd|›�«

tt««
}	|jt|«
«
}|j«D�	cgc]}	|	d‘Œ	}
}	tjdt!|
«
›d	|
d
d›t!|
«
dkDrdn
d›�«

|j#|
|«|j%«
Sc
c}	w#t&$rZ}tj)d
|›�«

tj)d|›�«

t+gdddd¬«cYd
}~|j%«
Sd
}~w
wxYw
#|j%«
wxYw
)zqSynchronous implementation of process_query. This is run in a thread by
        the async wrapper above.
        zUsing cached SQL: )
r&z```sqlÚz```zGenerated SQL: rzFound z fund IDs: Né
z...zSQL execution error: zFailed SQL: é
©ÚitemsÚtotalÚpageÚ	page_sizeÚtotal_pages)r/rÚloggerÚinfor!Úformat_messagesrÚinvokeÚcontentr,ÚreplaceÚnextrÚexecuterÚfetchallÚlenÚ_fetch_funds_by_idsÚcloseÚ	ExceptionÚerrorr
)r"r&r0Ú	cache_keyÚ	sql_queryÚmessagesÚresponseÚ
db_sessionÚresultÚrowÚfund_idsÚ
es            r#r4z"QueryProcessor._process_query_syncŠsç
€ð×'Ñ'¨Ó1ˆ	ð˜×(Ñ(Ñ(Ø×(Ñ(¨Ñ3ˆIÜ�K‰KÐ,¨Y¨KÐ8Õ9ð—‘×6Ñ6ÀÐ6ÓIˆHØ—x‘x—‘ xÓ0ˆHØ ×(Ñ(×.Ñ.Ó0ˆIð"×)Ñ)¨(°BÓ7×?Ñ?ÀÀrÓJ×PÑPÓRˆIð+4ˆD×Ñ˜YÑ'Ü�K‰K˜/¨)¨Ð5Ô6ôœ&›(“^ˆ
ð	Ø×'Ñ'¬¨Y«Ó8ˆFØ*0¯/©/Ó*;Ö< 3˜˜A›Ð<ˆHÐ<Ü�K‰KØœ˜X›˜ {°8¸C¸R°=°/Ì3ÈxË=Ð[]ÒK]Á%ÐceÐAfÐgô
ð×+Ñ+¨H°jÓAð
×ÑÕùò=øôò	Ü�L‰LÐ0°°Ð4Ô5Ü�L‰L˜<¨	 {Ð3Ô4ä$Ø 
¨
°RÀQôô
ð
×ÑÕûð	ûð
×ÑÕúsCÄ,FÄ0FÄ<AFÆFÆ	H
Æ'?G<Ç&
H
Ç'HÇ<H
È
HÈHrUc
óD—|
s#t
gd
d|
rt|
«
d
¬«Sdd
¬«Stt««
}	d}|�i|j	t
«
j
tt
j«
«
jt
j|k(«
j«}|j	t«
j
ttj«
jtj«
ttj«
jtj «
ttj«
jtj"«
ttj$«
ttj"«
«jtjj'|
«
«
j)«}g}|D�
]o}|j}d}	|�t+||d¬«}	|jdd	D�
cgc]#}
t-|
j|
j.¬
«‘Œ%}}
|j$r3dj1|j$D�cgc]}|j.‘Œc}«
n
d}
|j"r|j"dd	n
gD�cgc]#}t3|j|j.¬