GitHub - valkey-io/valkey: A flexible distributed key-value database that is optimized for caching and other realtime workloads. A flexible distributed key-value database that is optimized for caching and other realtime workloads. - valkey-io/valkey

@valkeyio.bsky.social - a fast #keyvalue #DB.
is wire-compatible with #Redis OSS 7.2.

#OpenSource BSD-3 license
github.com/valkey-io/va...

GitHub - NikolayS/pgque: PgQue – Zero-bloat Postgres queue. One SQL file to install, pg_cron to tick. PgQue – Zero-bloat Postgres queue. One SQL file to install, pg_cron to tick. - NikolayS/pgque

PgQue – Zero-bloat #Postgres #queue. One #SQL file to install. Requires pg_cron.
#PostgreSQL
#OpenSource Apache 2.0 lic
github.com/NikolayS/pgque

Hyperagents Self-improving AI systems aim to reduce reliance on human engineering by learning to improve their own learning and problem-solving processes. Existing approaches to self-improvement rely on fixed, ha...

⭐️Peper: arxiv.org/abs/2603.19461

GitHub - facebookresearch/HyperAgents: Self-referential self-improving agents that can optimize for any computable task Self-referential self-improving agents that can optimize for any computable task - facebookresearch/HyperAgents

#Meta: self-improving #coding #agents 🧠⚙️
Write code → run → evaluate → improve

Not just task performance—also the process of improving itself 🔁
Learns across runs + refines its own optimization loop

Meta-learning in practice 🚀
📜 Creative Commons lic
In the 💬👇🔗 to paper📝

#AI #LLM #Agents Repost 🔁

GitHub - apache/hugegraph: A graph database that supports more than 100+ billion data, high performance and scalability (Include OLTP Engine & REST-API & Backends) A graph database that supports more than 100+ billion data, high performance and scalability (Include OLTP Engine & REST-API & Backends) - apache/hugegraph

@apache.org HugeGraph full-stack graph system: #GraphDB, Computing, and #AI. Supports High Availability & container-ready.

Complete #graph data processing: storage, real-time querying, and offline analysis. Supports #Gremlin, #Cypher query languages, #REST #API, #SDK.

#OpenSource Apache 2.0 lic

GitHub - datalab-to/chandra: OCR model that handles complex tables, forms, handwriting with full layout. OCR model that handles complex tables, forms, handwriting with full layout. - datalab-to/chandra

ChandraOCR 2 - 4B model
#olmOCR benchmark 85.9%
#OCR #AI #LLM #SLM
#OpenSource Apache 2.0 lic
github.com/datalab-to/c...

linux/Documentation/process/coding-assistants.rst at master · torvalds/linux Linux kernel source tree. Contribute to torvalds/linux development by creating an account on GitHub.

⭐️GitHub full Guidelines
github.com/torvalds/lin...

@linuxkernel.bsky.social guide 4 coding with #AI #agents 🐧

-Follow standard dev+ submission rules
-Disclose tool+ model used
-No AI “Signed-off-by” (DCO= human only)
-Humans review & own the code

You can’t blame AI- it executed your request. Responsibility stays with you

#LLM #OpenSource Repost 🔁

GitHub - reaviz/reagraph: 🕸 WebGL Graph Visualizations for React. Maintained by @goodcodeus. 🕸 WebGL Graph Visualizations for React. Maintained by @goodcodeus. - reaviz/reagraph

⭐️GitHub github.com/reaviz/reagr...

#WebGL #Graph #Visualization in #React
#OpenSource Apache 2.0 lic
Link in first 💬 👇

⭐GitHub: github.com/run-llama/li...
By @llamaindex.bsky.social

Self-hosted #OpenSource document parser with build-in #Tesseract #OCR. Plugins: #EasyOCR, #PaddleOCR.
#CLI, #API or #TypeScript bindings.

#OSS Apache 2.0 lic
Link in the first 💬👇

I would suggest opening a GitHub issue, describing how you test.

GitHub - nanograph/nanograph: On-device property graph database. Schema-as-code. One CLI → One Folder. No Server. Think: DuckDB for graphs. On-device property graph database. Schema-as-code. One CLI → One Folder. No Server. Think: DuckDB for graphs. - nanograph/nanograph

On-device #CLI #GraphDB.
Search: Semantic, Fuzzy, #BM25, Hybrid, Graph-Constrained Reranking.

Executables: install via #Brew for #Linux or #macOS or compile Rust from Source.
#OpenSource MIT lic
#Agents #Database #Graph #VectorDB

GitHub - ddickmann/vllm-factory: Production inference for encoder models - ColBERT, GLiNER, ColPali, embeddings etc. - as vLLM plugins for online and in-process deployment Production inference for encoder models - ColBERT, GLiNER, ColPali, embeddings etc. - as vLLM plugins for online and in-process deployment - ddickmann/vllm-factory

Inference #vLLM plugins ⚙️ faster encoders, poolers, structured prediction, embeddings

#SLMs enable real value

🔎 Multi-vector ModernColBERT
🏷️ NER + entity linking GLiNER
🧾 Schema extraction
📊 Reranking
🖼️ Multimodal ColPali/ColQwen
💻 CPU/GPU local

#AI #LLM #RAG Repost 🔁
📜 Apache 2.0 lic #OpenSource

GitHub - blaxel-ai/sandbox Contribute to blaxel-ai/sandbox development by creating an account on GitHub.

#Sandbox by Blaxel- an isolated micro #VM.
Run generated by #agents code or #Harness agents themselves inside a secure environment.

#Python & #Typescript bindings
#OpenSource MIT lic
github.com/blaxel-ai/sa...

GitHub - timescale/pg_textsearch: PostgreSQL extension for BM25 relevance-ranked full-text search. Postgres OSS licensed. PostgreSQL extension for BM25 relevance-ranked full-text search. Postgres OSS licensed. - timescale/pg_textsearch

#BM25 search extension for PostgreSQL DB
Full Text Searxh #DataBase
#OpenSource #PostgreSQL #OSS license
github.com/timescale/pg...

GitHub - vercel-labs/just-bash: Bash for Agents Bash for Agents. Contribute to vercel-labs/just-bash development by creating an account on GitHub.

A #sandbox to execute #bash #scripts in a secure environment designed to be used by #agents. By @vercel.com

Also supports running #Python & #JavaScript
#Shell #AI #LLM
#OpenSource Apache 2.0 lic
github.com/vercel-labs/...

This is a prompt engineering framework. Not designed for robotics.
It generates “trajectories” aka variants of prompts & then “reflects”, selecting the best option LLM thinks. Unsupervised option however is not the most stable or predictable, for best results use labeled ground truth.

Flex inference  |  Gemini API  |  Google AI for Developers Learn how to optimize costs with the Flex inference tier

#Google #DeepMind released the Flex #API for all #Gemini #LLM models. 50% #costsreduction, designed for latency-tolerant workloads with 1-15 minutes turnaround.
#AI

ai.google.dev/gemini-api/d...

Container diagnostic tool. #Observability 4 #Linux kernel events, app-layer behavior, on demand, no prior config / installation. Monitor net, I/O, mem, syscalls, & high-level app events: HTTP, DNS, DB queries.

injects #eBPF programs at runtime to extract metrics. #K8s

#OpenSource Apache 2.0 lic

So back to caveman-style language: it’s only part of the solution. On average it’s likely better to use it than not. But for optimal results, you have to apply it only to problems where the LLM tends to overthink, leading to incorrect results.

4. Optimal LLM deployment **requires problem-
aware routing** with scale-specific prompting: you need a mechanism to identify problem types prone to overthinking and **apply brevity constraints selectively**

3. LLM is a great place to store knowledge, you just need to know how to query it to retrieve it.

According to this research,
1. Brevity is beneficial only with larger models.
2. Larger models require more careful prompt engineering to access their full capabilities depending on the model size.

GitHub - JuliusBrussee/caveman: 🪨 why use many token when few token do trick — Claude Code skill that cuts 65% of tokens by talking like caveman 🪨 why use many token when few token do trick — Claude Code skill that cuts 65% of tokens by talking like caveman - JuliusBrussee/caveman

Make your agent speak less. In a caveman-style language “why use many token when few do trick”

It only affects output tokens — thinking/reasoning tokens are untouched. “Caveman no make brain smaller. Make mouth smaller.”
#Agents #AI #LlM #PromptEngineering
#OpenSource MIT lic

GitHub - averdeny/graphglot: Parse, validate, transpile, and analyze graph query languages. Parse, validate, transpile, and analyze graph query languages. - averdeny/graphglot

Parse and validate #graph query written in #Cypher or #GQL before you send it to your #GraphDB. Useful for cases when an agent generates the query.
#Python Bindings, #CLI #Agents #DataBase
#OpenSource Apache 2.0 lic
github.com/averdeny/gra...

Install Wren AI as a #CLI tool (no containers) and query your DB from terminal or with UI! Has build-in @duckdb.org , supports #PostgreSQL, #BigQuery, #Snowflake, #MySQL #MariaDB, #Oracle DB, #Microsoft #SQL, Clickhouse, etc. #Python #DataBase #AI #LLM #Agents #RAG

#OpenSource AGPL 3.0 lic

RAG Made Simple: The Complete Visual Guide to Retrieval-Augmented Generation (Super AI Engineering Series) Amazon.com: RAG Made Simple: The Complete Visual Guide to Retrieval-Augmented Generation (Super AI Engineering Series) eBook : Diamant, Nir: Kindle Store

$1 #kindle #book about various #RAG techniques, including #GraphRAG and Evaluations. The #promo is available for 24h only!
#Graph #GraphDB #AI #LLM
www.amazon.com/RAG-Made-Sim...

Yep. When you treat LLM not as intelligence but as a DB, your approach shifts, now you know the goal is to figure out how to query it to get a better answer. The best way to do so is by optimizing prompt with algorithms like GEPA or ACE with labeled data.