Opiniated RAG for integrating GenAI in your apps 🧠 Focus on your product rather than the RAG. Easy integration in existing products with customisation! Any LLM: GPT4, Groq, Llama. Any Vectorstore: PGVector, Faiss. Any Files. Anyway you want.
EXPLAIN ANALYZE
WITH RECURSIVE knowledge_tree(id, file_name, url, extension, status, SOURCE, source_link, file_size, file_sha1, created_at, updated_at, metadata, is_folder, user_id, parent_id, sync_id, sync_file_id) AS
(SELECT knowledge_1.id AS id,
knowledge_1.file_name AS file_name,
knowledge_1.url AS url,
knowledge_1.extension AS extension,
knowledge_1.status AS status,
knowledge_1.source AS SOURCE,
knowledge_1.source_link AS source_link,
knowledge_1.file_size AS file_size,
knowledge_1.file_sha1 AS file_sha1,
knowledge_1.created_at AS created_at,
knowledge_1.updated_at AS updated_at,
knowledge_1.metadata AS metadata,
knowledge_1.is_folder AS is_folder,
knowledge_1.user_id AS user_id,
knowledge_1.parent_id AS parent_id,
knowledge_1.sync_id AS sync_id,
knowledge_1.sync_file_id AS sync_file_id
FROM knowledge AS knowledge_1
WHERE knowledge_1.parent_id = '40ba47d7-51b2-4b2a-9247-89e29619efb0'::UUID
UNION ALL SELECT knowledge_2.id AS id,
knowledge_2.file_name AS file_name,
knowledge_2.url AS url,
knowledge_2.extension AS extension,
knowledge_2.status AS status,
knowledge_2.source AS SOURCE,
knowledge_2.source_link AS source_link,
knowledge_2.file_size AS file_size,
knowledge_2.file_sha1 AS file_sha1,
knowledge_2.created_at AS created_at,
knowledge_2.updated_at AS updated_at,
knowledge_2.metadata AS metadata,
knowledge_2.is_folder AS is_folder,
knowledge_2.user_id AS user_id,
knowledge_2.parent_id AS parent_id,
knowledge_2.sync_id AS sync_id,
knowledge_2.sync_file_id AS sync_file_id
FROM knowledge AS knowledge_2
JOIN knowledge_tree ON knowledge_2.parent_id = knowledge_tree.id)
SELECT knowledge.id,
knowledge.file_name,
knowledge.url,
knowledge.extension,
knowledge.status,
knowledge.source,
knowledge.source_link,
knowledge.file_size,
knowledge.file_sha1,
knowledge.created_at,
knowledge.updated_at,
knowledge.metadata,
knowledge.is_folder,
knowledge.user_id,
knowledge.parent_id,
knowledge.sync_id,
knowledge.sync_file_id,
prompts_1.id AS id_1,
prompts_1.content,
prompts_1.title,
prompts_1.status AS status_1,
brains_1.brain_id,
brains_1.name,
brains_1.description,
brains_1.status AS status_2,
brains_1.model,
brains_1.max_tokens,
brains_1.temperature,
brains_1.last_update,
brains_1.brain_type,
brains_1.prompt_id,
syncs_1.id AS id_2,
syncs_1.name AS name_1,
syncs_1.provider,
syncs_1.email,
syncs_1.user_id AS user_id_1,
syncs_1.credentials,
syncs_1.state,
syncs_1.created_at AS created_at_1,
syncs_1.updated_at AS updated_at_1,
syncs_1.last_synced_at,
syncs_1.additional_data
FROM knowledge
JOIN knowledge_tree ON knowledge.id = knowledge_tree.id
LEFT OUTER JOIN (knowledge_brain AS knowledge_brain_1
JOIN brains AS brains_1 ON brains_1.brain_id = knowledge_brain_1.brain_id) ON knowledge.id = knowledge_brain_1.knowledge_id
LEFT OUTER JOIN prompts AS prompts_1 ON prompts_1.id = brains_1.prompt_id
LEFT OUTER JOIN syncs AS syncs_1 ON syncs_1.id = knowledge.sync_id
WITH RECURSIVE knowledge_tree AS (
SELECT *
FROM knowledge
WHERE parent_id = :parent_id
UNION ALL
SELECT k.*
FROM knowledge k
JOIN knowledge_tree kt ON k.parent_id = kt.id
)
SELECT * FROM knowledge_tree
linear[bot]
commented
1 month ago
Current generate query:
Should be simpler:
Previous implementation