Pydantic

2.6K posts

Pydantic banner
Pydantic

Pydantic

@pydantic

The Pydantic Stack: Pydantic Validation, Pydantic AI, Pydantic Logfire, Pydantic Evals, and Pydantic AI Gateway

London Katılım Ağustos 2022
167 Takip Edilen19.5K Takipçiler
Sabitlenmiş Tweet
Pydantic
Pydantic@pydantic·
AI observability can cost more than it should. We ran the numbers: same scale, same workload, four tools. At moderate production load, Logfire is up to 40x less expensive than the alternatives. Spans-based pricing. No proprietary units. AI native and full-stack observability in one platform. See the full comparison: pydantic.dev/articles/ai-ob…
Pydantic tweet media
English
10
7
66
25.4K
Pydantic
Pydantic@pydantic·
Your dashboard says your agent's average is OK. Your bill disagrees. We built Agents and LLMs views so you can see the runaway: the one in ten thousand that fired forty tools where the median fired three. Part 2 of Agents Week. Read the story: pydantic.dev/articles/logfi…
Pydantic tweet media
English
1
0
12
602
Pydantic
Pydantic@pydantic·
A good harness arms a single run. A loop of agents is what you build when one run isn't enough. It delegates to sub-agents, writes its own coordination, and writes new tools when it has none. The improvement unit is the loop. Read part 2 of 3 from @dasfacc on sub-agents, dynamic workflows, and agents that build their own tools: ordnl.link/oLJA4o9
Pydantic tweet media
English
0
2
24
1.5K
Pydantic retweetledi
Laura Summers
Laura Summers@summerscope·
Really looking forward to giving my talk “The Human-in-the-loop is tired” Wednesday 11:40 room S3A. It’s a fun one, hope to see you there ✨ ep2026.europython.eu/session/the-hu…
Pydantic@pydantic

The Pydantic team is heading to @EuroPython 2026 🐍 @summerscope and @marcelotryle are giving talks. @jirikuncar and @lais_bsc will be roaming the halls with stickers (supply is generous, ask away). If you use Pydantic, Pydantic AI AI, Logfire, Pydantic Evals, Monty, HTTPX2, or anything else we maintain, come say hi. We'd love to hear what you're building. Feedback, feature requests, and mild complaints all welcome.

English
0
1
5
676
Pydantic
Pydantic@pydantic·
The Pydantic team is heading to @EuroPython 2026 🐍 @summerscope and @marcelotryle are giving talks. @jirikuncar and @lais_bsc will be roaming the halls with stickers (supply is generous, ask away). If you use Pydantic, Pydantic AI AI, Logfire, Pydantic Evals, Monty, HTTPX2, or anything else we maintain, come say hi. We'd love to hear what you're building. Feedback, feature requests, and mild complaints all welcome.
Pydantic tweet media
English
1
3
19
2.6K
Pydantic
Pydantic@pydantic·
A perfect agent you never shipped is just an expensive opinion. Evals tell you an agent is right. Only production tells you it's the right agent. Agents Week starts today: ship rough, read the traces, and run agents like cattle, not pets. pydantic.dev/articles/agent…
English
1
1
12
2.1K
Pydantic retweetledi
Maziyar PANAHI
Maziyar PANAHI@MaziyarPanahi·
Full stack, all open: brain GLM-5.2 (@Zai_org) via @huggingface Inference Providers, eyes Qwen3-VL (@Alibaba_Qwen) running on Llama.cpp. Viewer: Mol*, open-source. Agent: @pydantic AI. Structure: EGFR + erlotinib from the RCSB PDB. 6 renders, one prompt, every critique real. 🤗
English
0
3
25
2.1K
Pydantic
Pydantic@pydantic·
@N_Rikhil We're trying! How should we spread the word better?
English
0
0
0
40
Remiscus
Remiscus@N_Rikhil·
@pydantic you guys gotta do better PR! day one here, so much better than langchain langgraph or whatever else
English
1
0
1
47
Pydantic
Pydantic@pydantic·
If your agent runs code, you've had to decide what to run it in. The default answer is often a full Linux VM, though most agents need a fraction of that. Join @samuelcolvin on Wed, July 29 for the case that a curated runtime beats a full sandbox, with a live demo of Pydantic Monty: pydantic.io/WgLDu
Pydantic tweet media
English
0
5
23
1.7K
Pydantic retweetledi
Mr. Ånand
Mr. Ånand@Astrodevil_·
Powered by @pydantic AI agent framework!
Mr. Ånand@Astrodevil_

Built a model battle playground for @nebiustf models with 3 modes: Code, Design, and Game. I put GLM-5.2 and Kimi K-2.7 Code to test on the same task. Rules: each model builds, reviews what it built, and gets 3 attempts to fix its own errors. The prompt was simple, but the task was complex to build. Both failed at building a proper game. ✅ Design Mode: → GLM-5.2: 15,768 tokens, $0.044, 71s → Kimi-K2.7: 12,045 tokens, $0.036, 140s ✅ Code Mode: → Kimi-K2.7: 11,776 tokens, $0.034, 103s → GLM-5.2: 11,562 tokens, $0.032, 234s Overall: GLM had better designs and game. Kimi was good at app logic. GLM was faster for design. Kimi was faster for code. Choose based on your workload from Token factory 🔥

English
0
3
7
996

Keşfet

@dasfacc @EuroPython @summerscope @marcelotryle @jirikuncar @lais_bsc @Zai_org @huggingface