Priyadharshini P

.

[Hrida AI @HridaAI]

Fireworks got breached. Here’s the part nobody’s talking about.

> Oct 6: stolen internal credentials. Attackers read environment variables from evaluation jobs and grabbed customer API keys.

> Oct 23: a third-party cloud provider hosting Fireworks’ R&D machines gets hit.

> Oct 24: attackers walk through that door into dev machines and run their own workload. Customer datasets and incomplete model weights are accessed.

The goal wasn’t to leak anything. It was LLMJacking: steal AI access, resell it through illicit routing gateways.

Three lessons:

> Your security includes everything you rent. The second breach didn’t start at Fireworks.
> Secrets in environment variables are one bad day from public. Short-lived, scoped credentials beat long-lived keys.
> Deleting a leaked key from your storage doesn’t kill it. Revoke it at the source.

We run an AI gateway, so we hold keys and see traffic. That makes us a target too. We build assuming we’re already breached, and the keys and data we hold are stored encrypted.

Rotate your AI keys today.

[Hrida AI @HridaAI]

Most LLM bills are high because one flagship model answers everything. 8 practical fixes engineering teams can ship this week to cut AI costs. https://t.co/dx7QZkSjwj

[Hrida AI @HridaAI]

🚨 LiteLLM, a Python package for connecting to AI model providers, has been flagged as compromised.

Version 1.107.0 is marked as malware on PyPI.

Cloud credentials. Developer secrets. Provider API keys.

All sitting in the same environment as your gateway.

If your team has been affected by today's LiteLLM security incident, we're happy to help.

Security incidents can happen to anyone, and we know how disruptive they can be.

At Spro, we offer a hosted AI gateway and AI security platform:

> One OpenAI-compatible API across multiple providers
> Routing and automatic failover
> PII, PHI and API key masking before data reaches a provider
> Prompt injection and jailbreak detection
> Per-request traces, cost controls and caching
> BYOK, so you keep your own provider keys

No gateway package to install. No gateway code to patch inside your infrastructure.

Most apps only need a base URL and key change to switch.

If you're exploring alternatives, reach out or send a DM.

We'll help make the transition easier.

https://t.co/JGGw6JHA5g

#litellm

[Hrida AI @HridaAI]

🚀 Claude Haiku 5.5 is now live on SPRO.

Anthropic’s latest Haiku is built for fast, high-volume, cost-sensitive AI workloads, while bringing stronger capabilities for coding, computer use, and agentic applications.

→ 1M context
→ Stronger coding & computer use
→ Multimodal capabilities
→ $0.10/M input · $0.50/M output

Access it through the same OpenAI-compatible API.

One API. 350+ models. One secure gateway.

Try it on SPRO: claude-haiku-5.5

#Claude #ClaudeHaiku #Anthropic #AI #LLM #GenAI #AIAgents #SPRO #HridaAI

[Hrida AI @HridaAI]

Why send your customer data to upstream AI models?

Your application sends a prompt.
Your AI Gateway routes it to the upstream model.
But your gateway should be doing more than just routing.

Your AI model needs the context.
It doesn't necessarily need the sensitive data.

That's where SPRO comes in. SPRO sits between your application and upstream AI models, inspecting every request before it leaves your environment.

When sensitive data is detected, SPRO can:

𝗟𝗢𝗚 - record it and let it through
𝗪𝗔𝗥𝗡 - flag it and continue
𝗥𝗘𝗗𝗔𝗖𝗧 - mask sensitive data with placeholders
𝗕𝗟𝗢𝗖𝗞 - stop the request entirely

No one-size-fits-all security policy.

Define what happens based on the data, the user, the model, or your organization's requirements. Your AI Gateway shouldn't just route requests.

𝗜𝘁 𝘀𝗵𝗼𝘂𝗹𝗱 𝗱𝗲𝗰𝗶𝗱𝗲 𝘄𝗵𝗮𝘁 𝗱𝗮𝘁𝗮 𝗴𝗲𝘁𝘀 𝘁𝗼 𝘁𝗵𝗲 𝗺𝗼𝗱𝗲𝗹.

SPRO | Security & Routing for AI. 🧵

[Priyadharshini P @Priya_dharshu24]

Hey devs 👋

Using GPT, Claude, Gemini, DeepSeek, Qwen or GLM? Managing different APIs and security for each one can get annoying.

SPRO to makes it simple - one gateway to access multiple AI models, with security guardrails built in.

Try it for free: https://t.co/Phyda8gi0k

[Hrida AI @HridaAI]

𝗚𝗿𝗼𝗸 4.7 is now available on 𝗦𝗣𝗥𝗢. 🚀

Built for:
• Coding & agentic tasks
• Long-running workflows
• Stronger self-verification
• 500K context
• $2/M input · $6/M output

Access Grok 4.7 through one unified API alongside 350+ models.

https://t.co/TzqkW6z0bP

[Hrida AI @HridaAI]

Why pay a percentage on top of your AI spend?

𝗦𝗣𝗥𝗢 𝗶𝘀 𝗻𝗼𝘄 0% 𝗺𝗮𝗿𝗸𝘂𝗽.

Hundreds of models. One API.
Routing. Fallbacks. Guardrails. Observability.
Enterprise controls.

You pay for the tokens.
𝗡𝗼 𝗲𝘅𝘁𝗿𝗮 𝗺𝗮𝗿𝗸𝘂𝗽.

Just a simple, flat platform fee.

[Hrida AI @HridaAI]

GLM-5.3-FlashX is live on SPRO.

- 1M-token context
- 320B total parameters, 18B active
- Native text + image understanding
- Tool calling + structured outputs
- Up to 200 tokens/s inference speed

Route GLM-5.3-FlashX + 195+ models through SPRO:

https://t.co/qHxQ0wczhO

[Hrida AI @HridaAI]

Your AI stack shouldn't need 5 API keys and zero guardrails.

Spro = 1 gateway → any model (OpenAI, Anthropic, Groq, more) + 29 safety guardrail checks on every request, out of the box.

BYOK or managed key. Your call.

We just launched on @ProductHunt 🎉

[Hrida AI @HridaAI]

Here’s the 60-second setup to switch from any AI provider to SPRO⚡

Base URL https://t.co/IWHJSbeonc

API Key = YOUR_SPRO_API_KEY

Works with @OpenAI SDK.

Change your base URL, API key, and model then send your first request. That’s it. Your AI traffic now runs through SPRO 🚀

[Priyadharshini P @Priya_dharshu24]

@RV_Smirnov We use a structured intermediate representation generated from the prompt rather than relying purely on vision models. Dexis understands intent, relationships, hierarchy, and flow first then generates diagrams from that structure

[Priyadharshini P @Priya_dharshu24]

@RomanGweb3 It works seamlessly with multiple input formats text, voice notes, PDFs, docs, and more. We’re also currently working on supporting video inputs.

[Priyadharshini P @Priya_dharshu24]

Dexis by Priyadharshini P https://t.co/A62rNkA48c

An AI tool that turns rough thoughts into structured, presentation-ready slides and diagrams in minutes.

Today we’ve launched Dexis on Peerlist 🚀
Would love to hear what the community thinks!
#AI #SaaS #BuildInPublic

[Priyadharshini P @Priya_dharshu24]

@neokd_20 2 ❤️‍🔥

[kuldeep @neokd_20]

@irabukht hey! would love to connect
https://t.co/XH9PzxjzLK

[Hrida AI @HridaAI]

🚀 Dexis Diagrams are now live.

Most system diagrams start the same way:

Open a diagram tool.
Drag boxes.
Connect arrows.
Rearrange everything again.
Hours just to explain one system.

What if architecture started with just a prompt? 👇

[Hrida AI @HridaAI]

Happy International Women’s Day 💜

Celebrate the dreamers and builders shaping the future

Get the Build Plan for $6.

Turn ideas, docs, videos & links into slides, diagrams & structured docs in seconds.

Offer ends March 10 → https://t.co/ShUA8czNAc

#InternationalWomensDay

[kuldeep @neokd_20]

Engineers hate drawing architecture diagrams.

So I’m building something for this in Dexis.
Shipping this weekend.

[Hrida AI @HridaAI]

Just like ideas, people, and innovations.
This Holi, may your days be filled with creativity, positivity, and vibrant new beginnings.

Wishing everyone a colorful and joyful Holi from @HridaAI ! 🌸
#HappyHoli #HridaAI #FestivalOfColors #Innovation #Startups