Notes from production AI.Not marketing content.
What we learn shipping AI onto client hardware: self-hosting economics, model choices, and security.
GraphRAG 3.3.0 has a graphrag update command, so rebuild-on-every-change is stale. LightRAG vs GraphRAG on local-model fit, Postgres storage and deletes.
Everything we have
written down.
Every post comes out of work we did: a client system, an internal build, or a benchmark we ran. Filter by topic above, or read straight through.
LightRAG vs GraphRAG (2026): Updates, Storage, Local Models
GraphRAG 3.3.0 has a graphrag update command, so rebuild-on-every-change is stale. LightRAG vs GraphRAG on local-model fit, Postgres storage and deletes.
OCT 09, 2026
Real Time Chunking in LeRobot: RTC vs Async Inference
LeRobot v0.6.1's async policy server never gives RTC its delay or previous chunk. Pick rtc or async by GPU location, then tune RTC from worst-case latency.
OCT 08, 2026
Open WebUI Entra ID SSO: Fix GUID Groups and Group Sync
Entra sends group GUIDs by default and drops the groups claim past 200 groups. Fix names, missing groups and wiped memberships in Open WebUI v0.11.4.
OCT 07, 2026
n8n Security Vulnerabilities: Which of 58 Hit Your Setup
n8n published 58 security advisories in six weeks, only 17 with a CVE. Map each to the config that exposes it, then run 1.123.83, 2.41.7 or 2.42.3.
OCT 06, 2026
LangGraph Checkpointer Postgres: PgBouncer and SSL Errors
At 3.1.2, the LangGraph Postgres checkpointer factory hardcodes prepare_threshold=0 and pipelines every write. Six errors traced, two undiagnosed.
OCT 05, 2026
OpenCode Offline Mode: Air-Gapped Setup for Local Models
OpenCode v1.18.34 has no offline flag. Its source shows 17 outbound paths, four with no setting at all, and the model catalog now calls models.opencode.ai.
OCT 02, 2026
vLLM Structured Output Not Working: HTTP 200, No JSON
vLLM removed guided_json in v0.12.0 but still answers it with HTTP 200 and free text. Ten failure modes stamped by version, their fixes, and a CI canary.
OCT 01, 2026
pg_search vs pg_textsearch: BM25 Hybrid Search in Postgres
pg_search vs pg_textsearch: both scored 0.68 nDCG@10 on SciFact, double native ts_rank_cd. Replicas, license and RLS decide which one you ship.
SEP 30, 2026
LeRobot n_action_steps vs chunk_size: How Many to Execute
SmolVLA's own ablation falls from 82.8% to 51.8% on LIBERO when it plays all 50 predicted actions. What n_action_steps and chunk_size do in LeRobot.
SEP 29, 2026Not ready to talk?
What we learn shipping AI systems, once a week. No announcements, no fluff.