PostgreSQL
PostgreSQL is an open-source object-relational database management system known for its high degree of SQL standards compliance, support for ACID transactions, and data integrity guarantees. It offers a rich set of data types, a powerful extension mechanism including custom functions and a plugin architecture, and high concurrency handling capabilities. It is widely used in enterprise applications, data analytics, and web services.
Core Facts
LangGraph's checkpoint store supports multiple backends by default, including in-memory, SQLite, and PostgreSQL.
The pgvector extension adds vector storage and similarity search capabilities to PostgreSQL, enabling Supabase to serve simultaneously as a traditional database and a vector database.
Supabase is an open-source alternative to Firebase that provides backend-as-a-service capabilities such as database, authentication, and real-time subscriptions based on PostgreSQL.
Timeline (last 90 days)
PostgreSQL's interactive shell is called psql, and its special commands begin with a backslash (such as \dt to list tables); MySQL's shell is called mysql and uses SQL-like commands such as show tables and describe.
The official Compose example notes that PostgreSQL 18 changed the data directory.
Historically, the maturation of open-source databases such as MySQL and PostgreSQL forced commercial databases to shift from per-seat licensing to usage-based or subscription pricing.
In the LangGraph multi-worker scenario, the developer hashes the thread ID into a 64-bit integer as the advisory lock key; a worker attempts to acquire the advisory lock before processing a thread, and on success it exclusively owns the conversation, while on failure it means another worker is already processing it.
PostgreSQL advisory locks come in two kinds, session-level and transaction-level: session-level locks are released automatically when the connection is closed, and transaction-level locks are released when the transaction ends.
The advantage of PostgreSQL advisory locks is that no extra table is needed and fault release is naturally achieved through the database connection lifecycle; the disadvantage is that they are bound to the database connection, so connection pool reuse or a network blip may cause the lock to be released unexpectedly, and they cannot directly express the semantics of a lease with a timeout, so in practice they are often used together with a lease table.
The non-chain part in production typically involves: authentication and authorization (Auth0, Clerk, or self-built JWT), application data (PostgreSQL or MySQL), vector storage (Pinecone, Qdrant, Weaviate, or pgvector), file storage (S3-compatible object storage), document parsing (Unstructured, LlamaParse, or a custom pipeline), usage tracking (custom event logs or analytics tools), and background execution (Celery, RabbitMQ, or cloud task queues).
The timings in the course show the native full-text search query at about 0.3 milliseconds and the BM25 query at about 9 milliseconds, but the author explicitly states this does not mean which is faster, because the former returns zero rows and the latter returns 10 rows.
BM25 requires global statistics (document frequency, average document length), so it usually needs support from a dedicated index structure, whereas PostgreSQL's built-in ts_rank lacks whole-database document frequency statistics.
pgTextSearch is an open-source extension under the PostgreSQL license that provides BM25 ranking; the query operator is <@>, which returns the negative of the BM25 score, so sorting must be done in ascending order.
40 more timeline events
All Facts (20)
LangGraph's checkpoint store supports multiple backends by default, including in-memory, SQLite, and PostgreSQL.
90%VerifiedThe pgvector extension adds vector storage and similarity search capabilities to PostgreSQL, enabling Supabase to serve simultaneously as a traditional database and a vector database.
80%VerifiedSupabase is an open-source alternative to Firebase that provides backend-as-a-service capabilities such as database, authentication, and real-time subscriptions based on PostgreSQL.
80%VerifiedDify's core architecture adopts a microservices design, using Docker Compose to decouple and deploy components such as the API service, Worker, Web frontend, PostgreSQL, Redis, and vector database.
80%VerifiedAfter switching to Plan B, scene loading time was reduced from 3.3 seconds to approximately 0.5 seconds, approximately 6.8 times faster.
75%VerifiedLakebase is a new type of database form proposed within the Databricks ecosystem, whose core concept is to build Postgres-compatible transactional database capabilities on top of a data lakehouse architecture.
75%Verifiedpgvector is an open-source extension for PostgreSQL that natively adds vector data types and similarity search capabilities to the relational database, supporting both exact nearest-neighbor and approximate nearest-neighbor (HNSW, IVFFlat index) retrieval modes.
70%VerifiedRLS（Row Level Security，行级安全策略）是PostgreSQL原生特性，允许在数据库层面定义精细的访问控制规则
70%VerifiedSupabase is built on PostgreSQL, and RLS is a core security feature of the PostgreSQL database
70%VerifiedThe costs of storing plan configurations in a database include: data inconsistencies that can easily arise during product evolution, more combinations to test, and frequent reads potentially slowing down the application
65%VerifiedPostgreSQL can support vector retrieval via pgvector.
65%VerifiedThe principle of WAL (Write-Ahead Logging) is that all modification operations must be sequentially appended to the WAL log file before being written to the actual data file, used to achieve transaction durability and crash recovery.
65%VerifiedLangGraph's state is a TypedDict, and its checkpointing mechanism persists state snapshots to a storage backend (such as SQLite or PostgreSQL), supporting breakpoint recovery, time-travel debugging, and human-in-the-loop collaboration.
65%VerifiedLangfuse includes components such as a Web interface, PostgreSQL, Redis, ClickHouse, and a Langfuse Worker responsible for asynchronous evaluation.
65%VerifiedLangGraph serializes and persists the complete state after each node execution to external storage (such as PostgreSQL or Redis) via a checkpointing mechanism, enabling precise resumption from the point of interruption.
65%VerifiedEach PostgreSQL backend process has a base memory overhead of approximately 5–10 MB.
65%VerifiedPostgreSQL extensions are fully equal to the kernel at the query planner level, and the optimizer can perceive and utilize statistical information from extension indexes to generate execution plans.
65%VerifiedDjango's ORM supports multiple database backends, including PostgreSQL, MySQL, and SQLite.
65%VerifiedLangGraph's long-term storage supports multiple backends, including PostgreSQL for structured data, vector databases such as Pinecone and Chroma for semantic retrieval, and Redis for high-speed caching.
65%Partially VerifiedModern databases such as MySQL InnoDB and PostgreSQL both implement a buffer pool mechanism internally, managing cache replacement strategies through algorithms such as LRU.
100%