标签：#base

Envelope encryption con clave por usuario, qué se cifra y qué no, cómo lo puso a prueba una auditoría externa, y el incidente de rendimiento que provocó nuestro propio hardening de seguridad. Montaste tu app con IA rápido: le pides unos datos al usuario, llamas al modelo, guardas el resultado en la base de datos y a producción. Cómodo, sin complicaciones. Hasta que un día miras bien qué estás guardando. En Cuentopia generamos cuentos personalizados para niños. Para personalizar, un padre nos cuenta cómo es su peque: su carácter, qué le da miedo, qué está pasando en casa. El modelo no improvisa sobre la marcha: se apoya en un marco de criterios clínicos y pedagógicos para decidir cómo abordar cada situación, y luego lo reescribe todo en prosa. Visto de golpe, lo que teníamos en la base de datos era el diario emocional de un montón de menores. El RGPD lo trata como categoría especialmente protegida. El sentido común, también. ¿Y si se filtra la base de datos? ¿Y un backup mal guardado? ¿Y un acceso indebido con privilegios de admin? Relájate —bueno, primero asústate un poco; luego relájate—. Te voy a contar cómo pensamos el cifrado en reposo en serio: una arquitectura de tipo envelope encryption , con una clave maestra que no sale nunca de Cloud KMS (Google Cloud) y una clave por usuario que cifra los campos sensibles antes de que toquen la base de datos. Un aviso antes de seguir: te cuento el criterio y las decisiones, no el plano. No vas a encontrar aquí nombres de recursos, rutas de repositorio, ni el detalle exacto que le serviría de receta a alguien con ganas de probar suerte con nuestros datos. Y porque la seguridad honesta se cuenta entera, también te cuento dónde decidimos no llegar y por qué. ✨ Promesa: al terminar vas a entender, con criterio real de producto, cómo una familia sin ser expertos en cripto se planteó cifrar datos de menores — y por qué ciertas decisiones muy concretas no se hacen públicas nunca, ni en el artículo más honesto. El mapa Lo constru

Opinion: Your Tests Can't See What a Migration Destroys — Dry-Run It on a Clone A green test suite is the wrong tool for judging an AI-generated migration, because tests run against the post-migration schema and never observe the intermediate states where data disappears. The up migration is the visible artifact that gets reviewed, while the down migration is treated as an afterthought even though it is the only safety net when the deployment goes wrong. Free model access makes the problem structural: generation cost drops to zero, so migration volume rises, and every additional migration multiplies the surface for unreviewed data loss. Disclosure: This article was prepared as part of MonkeyCode's product outreach. Tests validate the destination, not the journey When a test suite runs against a migrated database, it confirms that the application can read the new schema, but it cannot confirm that the migration preserved the data it was supposed to preserve. The test runner connects after the migration has executed, so it never sees the moment when a column is dropped, a table is renamed, or a constraint is silently relaxed. A migration that passes every test can still destroy production data, because the tests were designed to validate application behavior, not migration safety. The standard mitigation is a staging database, but staging is a poor substitute for a dry run because it has different data, different volume, and different usage patterns. The dry run I recommend uses a clone of the production schema with a representative data sample, and it exercises both directions of the migration with data integrity checks at every step. The clone does not need to be large; a few thousand rows per table is enough to expose most destructive patterns. The dry-run workflow in five steps The workflow is deliberately mechanical, because the goal is to remove judgment from the verification process and reserve human attention for the migration's intent: Clone the schema and lo

By Michael Nocito , data analyst · Published August 8, 2026 By the end of this page you will be running real SQL against a real database with nothing installed, and you will know which of the free browser tools suits which job. You will also know the thing none of them puts on the front page: some of them run entirely inside your browser, and some upload whatever you paste to a stranger's server. That difference decides what you are allowed to practise on. Here is what to actually do today. If you want a database already loaded and questions already written, open sql-practice.com . If you want to create your own tables and share the result with someone, open DB Fiddle . Both start working immediately with no account. The short version: browser-only tools keep your data on your machine, server-backed tools do not, and neither kind is the right place for anything from work. Where the data goes is the one idea that should drive your choice, so it gets the picture. The original carries a diagram here. In words: Two panels side by side, each drawn as a laptop outline containing a browser window. In the left panel a small data box sits inside the browser window, with a short circular arrow looping back into itself, showing the data never leaves the laptop. In the right panel the same data box has a long arrow leading out of the laptop, across a gap, and into a separate server rack drawn beyond the laptop's edge, with a copy of the data box now sitting in the rack as well. The original box remains, showing the data has been copied out rather than moved. Every tool below was opened and checked on 8 August 2026. These sites change often, so the descriptions describe what was actually on screen, and anything I could not confirm by looking is not claimed here. 1. Run your first query, right now Before the explanation: what do you think has to exist on your computer for a SELECT statement to return rows? The honest answer is nothing at all, and that surprises people who have sp

By Michael Nocito , data analyst · Published August 8, 2026 By the end of this page you will have a real database sitting on your own computer, with 11 tables, 3,503 tracks and 412 customer invoices in it, and you will have run a query that proves every table arrived intact. Then you will run a join across two of those tables, which is the thing a single spreadsheet can never teach you. It takes about five minutes and costs nothing. Here is what to actually do today. Download the Chinook database file, open it in DB Browser for SQLite, and run one query that counts the rows in every table. If the counts match the ones printed below, you have a working practice environment and you can stop shopping for one. The short version: get Chinook_Sqlite.sqlite , open it, count the rows, then join two tables. Northwind and Sakila are the other two names you will see, and there is a table further down saying when each is the right pick. The reason a sample database beats the CSV you already have is one idea, so it gets the picture. The original carries a diagram here. In words: Two panels side by side. The left panel holds a single grid of rows and columns, standing alone with nothing attached to it. The right panel holds four smaller grids arranged around each other. A highlighted column at the edge of each small grid is joined by a solid line to a matching highlighted column on a neighbouring grid, so all four grids are wired together into a connected shape. The left panel has no lines at all, because there is nothing for a line to reach. Every number on this page is real. I downloaded Chinook v1.4.5 and Northwind on 8 August 2026 and ran each query with SQLite 3.51.1. The counts, the outputs and the row multiplication are what came back, not what should have come back. If you have no database software at all yet, how to set up a SQL database is the fifteen-minute version of that step, and this page picks up right after it. 1. Why one CSV is not enough Before the explanation:

Mark (aka Winsaucerer) here to show you how you can test your PostgreSQL database like a sorcerer. We are going to be using Spawn, a SQL build system supporting migrations and testing. You do not need to be using Spawn for migrations in order to use it for testing. Spawn does not require any extension installed. All you need is the spawn CLI and a psql connection to the database for Spawn to connect through. Spawn was built to solve some migration pains I've experienced, but I happily discovered that when used for testing, it is very powerful. To show you some of that power, we're going to use a contrived database example. It uses golden file testing to determine success. When the test runs, we capture the stdout and stderr output from psql, and compare that to expected output. Testing with Spawn involves these steps: Create a new test with spawn test new <name> and fill out the test steps Check test outputs with spawn test run <name> (or view the SQL that will be sent to psql via spawn test build <name> ) When outputs are as expected, create the golden file with spawn test expect <name> Run the test and compare to expected output with spawn test compare <name> For now, Spawn only supports connecting via psql, which means that you have access to all the features that psql provides. To get started, follow the Spawn install instructions: Install Spawn And then create a new folder on your system, and initialise a new project with a docker compose config ready for us to play with: # inside your new folder: spawn init --docker docker compose up -d You now have a running docker based PostgreSQL database and a spawn.toml file configured to connect to it. We are not assuming that you are using Spawn or any other tool for migrations, so you can manually create and update the database by connecting directly using psql: docker exec -ti postgres-db psql -U postgres Create the database ⚠️ Caution This post is not intended as an example of how to build an orders database. The des

Top Vector Databases for AI Agents in 2026: Qdrant vs Pinecone vs Weaviate vs PgVector vs Milvus Persistent memory is the foundation that turns a stateless LLM into a continuously improving, autonomous agent. In 2026, selecting a vector database is no longer just about raw Approximate Nearest Neighbor (ANN) speed. For AI agents, the critical requirements have shifted to: Payload & Metadata Filtering : Can you filter by tenant_id , user_id , and timestamp during vector graph traversal without sacrificing recall? Hybrid Search (BM25 + Dense Vectors + Sparse SPLADE) : Combining exact keyword matching (for code symbols and error codes) with semantic understanding. Multi-Tenancy & Memory Namespacing : Safely isolating memory blocks across thousands of users and sessions. Billion-Scale Quantization (Product Quantization & Scalar Quantization) : Slashing RAM costs by 75–90% in production. This guide provides a comprehensive architectural comparison of the top 5 vector databases for AI agents in 2026. Head-to-Head Comparison Matrix Feature / Metric Qdrant Pinecone (Serverless) Weaviate PgVector (PostgreSQL) Milvus Primary Architecture Rust-native, disk-backed Fully managed serverless Go-native, modular RAG PostgreSQL extension Distributed cloud-native Open Source Yes (Apache 2.0) Proprietary SaaS Yes (BSD-3) Yes (Open Source) Yes (Apache 2.0) Payload Filtering Exceptional (HNSW custom payload indexing) Good (Metadata filtering) Strong (Inverted index + HNSW) SQL WHERE clause Strong (Partition keys) Hybrid Search Native (Dense + Sparse vectors) Native hybrid Native BM25 + Vector SQL text search + pgvector Native multi-vector Quantization Scalar & Product Quantization (Binary) Automatic serverless compression PQ, BQ, SQ Halfvec, Binary Quantization Scalar / Product Quantization Best Fit High-performance agent memory & self-hosted RAG Zero-maintenance cloud SaaS GraphQL & multi-modal search Unified relational + vector apps Ultra-large enterprise (100M+ vectors) 1. Qdrant: The

Disclosure: I build Schemity , a desktop ERD tool - this post is from our blog and uses it for the examples. TL;DR: The database has a built-in place to document a column - COMMENT ON COLUMN in PostgreSQL, the COMMENT attribute in MySQL - and almost nobody fills it in, because a sentence of prose has to travel the same path as a schema change: a migration file, a review, a deploy. Schemity keeps field descriptions in the diagram instead, where editing one generates no SQL, reads existing database comments in on import, and exports the result as a data dictionary. You can document a database column without touching the database: write the description in the model rather than in the schema. That sounds like a dodge until you price the alternative. The database's own mechanism for column documentation, COMMENT ON COLUMN in PostgreSQL and the COMMENT attribute in MySQL, sends a sentence of prose down exactly the same path as a change to how data is stored - a migration file, a code review, an approval, a deploy window - and on MySQL it does something worse than that. Schemity keeps field descriptions in the diagram, where editing one produces no SQL at all. This is why so many production schemas have thousands of columns and almost no comments. Not because nobody wanted to write them. Because writing one costs a deploy. How do I document a database column without running a migration? Keep the description in the model rather than in the storage engine. A field description is a fact about what the column means to your team; it changes no type, no constraint, no index, and nothing about what the database will accept. When it lives in the diagram, editing it is like editing a comment in a code file: you change it, review it in the same pull request as everything else, and nothing has to run against production for it to take effect. The moment that description is a column comment, it stops being prose and becomes DDL. Now it needs a migration file, and the migration needs a

I Gave Five Graph Databases 256MB of RAM Each. Here's What Broke. CognoDB Cloud's free tier gives you a graph database instance with half a CPU core and 256MB of RAM. That's not a lot. It's also, honestly, a pretty realistic starting point a lot of real side projects and early-stage products live exactly there, on whatever the free tier happens to give them, and find out the hard way what their database does under pressure. So I decided to actually find out. I took CognoDB and lined it up against four other graph databases Neo4j AuraDB, FalkorDB, and ArangoDB gave every single one of them the same tiny resource budget, threw the same 198,050-edge dataset at all of them, and ran the same queries. No cherry-picking, no "best case" numbers. Just: here's a small VM's worth of resources, go. One of the databases I originally planned to include never even made it into the results. It crashed on startup. Not "slow to start" a full segfault, reproducibly, across two different versions, with nothing I threw at it fixing it. More on that below, because it's honestly one of the more interesting parts of this whole thing. The setup, quickly Five candidates going in: CognoDB (mandatory, since that's the actual point of this), Neo4j AuraDB Free, Memgraph, FalkorDB, and ArangoDB. Same dataset for all of them a real social-graph-shaped dataset from Stanford's SNAP collection, ~18.7k nodes and ~198k edges, sized specifically to fit inside every platform's free tier without anyone getting an unfair advantage. Same queries too: I wrote every single query 1-hop, 2-hop, 3-hop traversals, point lookups, filtered lookups, aggregations exactly once, then translated each one into whatever query language a given platform actually speaks. No platform ever got a "friendlier" version of a query than another. And everyone ran under the same 0.5 vCPU / 256MB RAM ceiling, whether that was their real cloud free tier or a Docker container I capped by hand to match. The one that didn't survive Memgra

There's a moment every developer hits the first time they connect an AI assistant to a real database: it works beautifully, the model writes a clean SELECT , you get your answer in seconds — and then a small, cold thought arrives. What if it had written DELETE instead? That worry is healthy. An AI agent that can query your production database is also, by default, an AI agent that can UPDATE , DROP , and TRUNCATE it. Large language models are probabilistic. They hallucinate. They misread a vague prompt like "clean up the test users" as an instruction to actually delete rows. You don't want the only thing standing between a confused model and your orders table to be good intentions. The fix isn't to keep AI away from your data. It's to make write operations structurally impossible — read-only by design, enforced at layers the model can't talk its way past. This post walks through how to do that properly, from the database grant all the way up to query-level guardrails. Why "just prompt it to be careful" fails The tempting shortcut is to add "only run SELECT queries, never modify data" to your system prompt and call it a day. Don't rely on this. Prompt instructions are suggestions, not enforcement. A cleverly worded user request, an injected instruction hidden in some data the model reads, or a plain misunderstanding can all lead the model to generate a destructive statement anyway. Real read-only access is enforced below the model — in places where no amount of clever text can override it. Think of it as defense in depth, with at least three independent layers: Layer What it stops Enforced by Database permissions Any write reaching the engine SQL GRANT / REVOKE Connection / replica Writes even being routed to a writable node Read replica, read-only transaction Query parser / broker Non-SELECT statements before they run SQL parsing, allowlists Any one of these is decent. All three together mean a write has to defeat your database engine, your routing, and your parser s

The database platform Harper advocates for a single-runtime architecture that keeps application code and data together, with its benchmark against a Vercel-based stack reporting significantly better performance on live, personalized-data workloads. Harper recently released version 5.2, with a new record cache and more throughput per node. By Renato Losio

Last week I turned my data audit into a build step : a check that runs before anything else and fails the build when the database and any static copy of my travel site's legal-status data disagree. It ended the era of the site contradicting itself. It did nothing about the site agreeing with itself on something false. That's not a hypothetical. The most expensive error the whole project found was a country whose law changed in January while every copy on my site — database, data files, search index — kept saying the old thing in perfect unison. Internal consistency was the camouflage . No diff between my own sources could ever have caught it, because every internal source was equally behind the world. A build gate proves agreement. Agreement is not truth. Something has to look outside. You can't diff against the world, but you can sample it The naive version of "look outside" is another audit — a human session checking primary sources jurisdiction by jurisdiction. I've done three of those now, and I know exactly what they're worth: they're correct the day they ship and they decay from that morning on. Laws don't change on my audit schedule. So the outside check became what the inside check became: a scheduled job. Once a week, a script asks a web-connected model — one that searches and cites, not one answering from training memory — for the current legal status of about fourteen jurisdictions, and compares each answer to the corresponding database row. Fourteen, not all 271, because the selection is doing the real work: A hot list is checked every single run: the highest-traffic pages plus the jurisdictions with active legislative motion — the places where being a month stale costs the most. Everything else sits on a rotating cursor : eight per run, round-robin, so every row on the site gets sampled roughly twice a year without any run costing more than a few cents. The whole thing runs on about seven cents a week. Two rules were non-negotiable, both inherited from

@Service public class SimpleUserService { @Autowired private JooqTemplate jt ; public List < user > selectUserInDept ( UserParam param ) { //If deptIDs==null or deptIDs. isEmpty automatically ignores this query condition // SELECT * FROM user_table WHERE name LIKE '%?%' AND dept_id IN (?,?...); return jt . queryv ( "user_table" , User . class , "name%" , param . getName (), "dept_id:in" , param . getDeptIds ()); } public List < user > selectUserNotInDept ( UserParam param ) { // SELECT * FROM user_table WHERE name LIKE '%?%' AND dept_id NOT IN (?,?...); return jt . queryv ( "user_table" , User . class , "name%" , param . getName (), "dept_id:notin" , param . getDeptIds ()); } }

Companies including Tesla and Base Power are vying for a piece of the rapidly growing market for home batteries. One technology has made it all possible.

ICE collected nearly a million DNA samples last year.

In my last post I wrote about an error that cost me a full evening: my pom.xml had the MongoDB starter, but my code was still full of JPA annotations. The compiler kept saying cannot find symbol: class Entity . That post was about the error. This post is about the fix — every single line I had to change to move my Task Manager project from MySQL to MongoDB. If you are planning the same switch, this is the checklist I wish I had. 1. The dependency Before (MySQL + JPA): <dependency> <groupId> org.springframework.boot </groupId> <artifactId> spring-boot-starter-data-jpa </artifactId> </dependency> <dependency> <groupId> com.mysql </groupId> <artifactId> mysql-connector-j </artifactId> <scope> runtime </scope> </dependency> After (MongoDB): <dependency> <groupId> org.springframework.boot </groupId> <artifactId> spring-boot-starter-data-mongodb </artifactId> </dependency> One starter replaces two dependencies. And this is exactly where my problem started — I added the new one but never removed the old one, so half my code still compiled and half did not. Remove the JPA starter completely. If you leave it in, the jakarta.persistence annotations still resolve, and you will not notice you are mixing two worlds until something breaks at runtime. 2. application.properties Before: spring.datasource.url = jdbc:mysql://localhost:3306/taskmanager spring.datasource.username = root spring.datasource.password = yourpassword spring.jpa.hibernate.ddl-auto = update spring.jpa.show-sql = true After: spring.data.mongodb.uri = mongodb://localhost:27017/taskmanager Five lines became one. No ddl-auto because MongoDB has no schema to create. No dialect because there is no SQL being generated. The database and the collection are created automatically the first time you insert a document. 3. The model class This is where most of the work was. Here is my actual Task class after the migration: package com.taskmanager.task_manager ; import com.fasterxml.jackson.annotation.JsonIgnore ; import org.

If you build with Prisma, you already know Prisma Studio. Run one command and you get a clean, visual way to browse and edit rows in your database. It's genuinely useful, and I reach for it every day while developing. But somewhere between "I need to look at my data" and "I need to let a support agent safely edit a customer's record in production," Prisma Studio quietly stops being the right tool. It was never trying to be that tool. It's a database viewer. An admin panel is something else, and the gap between the two is exactly the part that matters once real people and real permissions are involved. I ended up building a small package to fill that gap for my own Express + Prisma apps. Writing it forced me to be precise about what an admin panel actually adds on top of a database browser. Here's the distinction as I now understand it. A database browser shows rows. An admin panel governs them. Prisma Studio connects to your database and shows you everything. That's the point of it, and it's also why you'd never hand it to a non-engineer or expose it in production. It has no concept of who is looking, what they're allowed to do, or which rows they're allowed to touch. An admin panel's whole job is those three questions. The package I built mounts a React UI at /admin and a guarded JSON API under /admin/api/* on your existing Express app. Every single request through that API runs the same pipeline, in the same order: authentication → permission check → tenant scope → validation → Prisma mutation/query → optional audit event That ordering is the entire difference. A database browser skips straight to the mutation. An admin panel refuses to run the mutation until it knows the request is authenticated, permitted, scoped to the right tenant, and valid. Permissions and scope are two different questions This was the design decision I care most about, because collapsing these two into one is how data leaks happen. Permissions decide which actions a role may take. Can an ed

I got an AWS question and implemented it to make sure that the option is correct. A critical financial application runs on RDS for PostgreSQL. The requirements are tight: 1-second RPO, 60-second RTO, and database credentials rotated every 30 days without taking the application offline. Two independent problems. Two independent solutions. Prerequisites Check these before running terraform apply : RDS Proxy availability RDS Proxy is not available on all instance types. It requires instances with at least 2 vCPUs. db.t3.micro is not supported. db.t3.medium and above work. Terraform executor permissions The IAM principal running Terraform needs, at minimum: rds:CreateDBInstance rds:CreateDBProxy rds:CreateDBProxyTargetGroup rds:RegisterDBProxyTargets rds:ModifyDBInstance iam:CreateRole iam:AttachRolePolicy iam:PutRolePolicy iam:PassRole secretsmanager:CreateSecret secretsmanager:PutSecretValue secretsmanager:RotateSecret lambda:CreateFunction lambda:AddPermission ec2:CreateSecurityGroup ec2:AuthorizeSecurityGroupIngress ec2:CreateDBSubnetGroup AdministratorAccess on the account covers all of these. Lock it down after the initial setup. VPC requirements RDS Proxy runs inside your VPC. You need at least two private subnets in different Availability Zones. The rotation Lambda also runs inside the VPC so it can reach the RDS instance directly during the credential update step. The problem Database failure recovery RPO of 1 second means almost no data loss is acceptable. RTO of 60 seconds means the application must resume within a minute of a failure. A standard single-instance RDS setup fails both requirements: there is no automatic failover, and restoring from a backup takes far longer than 60 seconds. Credential rotation Rotating credentials on a schedule sounds simple until you factor in application downtime. If you update a password and the application still holds connections authenticated with the old one, those connections fail. The rotation mechanism needs to handle

In this article, we'll explore the mechanisms to solve the coordination problem. 8. Introducing Leases To address the problem of permanent ownership, distributed systems typically replace it with temporary ownership. This concept is known as a lease . Instead of granting indefinite control over a resource, the coordination service assigns ownership for a limited period of time. Rather than stating, “You own this resource until you explicitly release it,” the system instead says, “You own this resource for the next 30 seconds.” This changes the interaction model significantly. Acquire Lease | v Execute Work | v Renew Lease | v Continue Processing As long as the application remains healthy, it periodically renews the lease to maintain ownership. If the application crashes or becomes unresponsive, it can no longer renew the lease. Once the lease duration expires, ownership is automatically revoked. At that point, another application becomes eligible to acquire the lease and continue the work. Leases solve a critical problem in distributed systems: they prevent abandoned locks from blocking progress indefinitely . The system can recover automatically without manual intervention. However, while leases improve availability, they also introduce a new class of subtle and more complex problems. Leases Depend on Time To understand the next challenge, assume the lease duration is thirty seconds. Application A successfully acquires the lease. Lease Granted Duration = 30 seconds After twenty seconds, the JVM begins a long Full Garbage Collection cycle. This pause lasts forty seconds, significantly longer than the lease duration. The timeline now becomes problematic. Lease Granted | | Processing | | GC Pause (40 sec) | | Lease Expires While Application A is paused, the lease expires. During this time, another application requests access to the same resource. The coordination service observes that the previous lease has expired and therefore grants ownership to Application B. Appl

Introduction Your program keeps opening transactions. A signup writes a new user row. A checkout debits one account and credits another. A form submission updates three related tables at once. You wrap it all in BEGIN and COMMIT and move on, trusting that the database will handle whatever happens in between. Most of the time it does. But what is it actually promising you when it handles that? And what does it have to do behind the scenes to keep that promise? Say a user transfers ₹1,000 from Account A to Account B. The application runs two updates: subtract 1,000 from A, add 1,000 to B. Now say the server crashes right after the first update runs but before the second one does. Account A: -₹1,000 Account B: +₹0 That money didn't move. It vanished. No error message fixes that, and no user accepts "the server restarted" as an explanation for their missing balance. This is the exact problem a set of guarantees called ACID was built to solve. Most developers can recite the acronym, Atomicity, Consistency, Isolation, Durability, without being able to explain what any of the four words actually promise, or what the database has to do internally to keep those promises. This article tries to fix that. -- 1. What Is a Transaction? Before ACID makes sense, you need to understand what a transaction actually is. A transaction is a group of one or more database operations treated as a single logical unit of work. Either the whole group succeeds, or none of it does. The bank transfer above is a textbook transaction: two updates that only make sense together. In SQL, a transaction usually looks like this: BEGIN ; UPDATE accounts SET balance = balance - 1000 WHERE id = 1 ; UPDATE accounts SET balance = balance + 1000 WHERE id = 2 ; COMMIT ; BEGIN tells the database "everything from here on is one unit." COMMIT tells it "we're done, make it permanent." If something goes wrong in between, a constraint violation, a crash, the application deciding to cancel, the database can issue a RO

This is the data appendix for Posts 1–4 . The narrative and takeaways live in the main posts. This page is pure measurement. Eval set: 22 questions (P1) → expanded to 31 questions (P2 onward) Corpus evolution: P1/P2: ~3,700 docs (raw threads + code chunks) P3/P4: 16,315 docs (distilled threads + bursts + code) Quick comparison (same 31-question set) Metric Vector P2 Hybrid P2 Vector P3 Hybrid P3 Hybrid + Rerank (P4) recall@1 0.68 0.61 0.52 0.39 0.87 recall@3 0.84 0.65 0.71 0.65 0.94 recall@10 0.90 0.90 0.81 0.94 0.94 MRR 0.77 0.67 0.63 0.57 0.90 Takeaway: Hybrid alone never beat pure vector on this corpus. Hybrid + LLM rerank is the first clear win. P1 — Naive vector baseline Corpus: 3,000 raw issue threads + 687 code chunks Embeddings: BGE-M3 (1024d), max_seq_length=1024, HNSW cosine Numbers (22 questions) Metric Score recall@10 1.00 (22/22) recall@3 0.95 recall@1 0.77 (17/22) Main k=1 misses Exact error pastes ( TypeError: Object of type int64... , AttributeError: 'Depends'... ) — ranked 4–5 instead of 1 jsonable_encoder code chunk outranked by issues about the function API key header implementation (code vs similar issues) Paraphrase questions (dependency injection outside routes, custom 404) Pattern: Dense search is strong on recall@10 but weak when the query has a sharp lexical signal. Ops notes Ingest wall time ~40 min (GitHub API is the bottleneck) BGE-M3 OOM on Apple Silicon fixed by capping max_seq_length=1024 Python 3.13 + uv editable install issue fixed by pinning 3.12 P2 — Hybrid (vector + FTS + RRF) Corpus: Same size as P1, with better comment pagination and symbol-based code IDs Eval set: Expanded to 31 questions (added exact error pastes + rare identifiers) Numbers Metric Vector FTS Hybrid recall@1 0.68 0.42 0.61 recall@3 0.84 0.48 0.65 recall@10 0.90 0.65 0.90 MRR 0.77 0.47 0.67 Headline: Hybrid is not a strict win over vector-only. Where hybrid helped Exact error pastes (e.g. TypeError: int64 is not JSON serializable ) → moved from rank 5 → 1 Near-d