Synaphos AI

Synaphos is now live run your own AI workspace in minutes. No cloud. No API key. No subscription. Just install Ollama on your machine or server, pull a model, and connect it to Synaphos. Your AI runs entirely on your own hardware every inference stays on your device. Install Ollama → curl -fsSL ollama.com/install.sh | sh Pull a model → ollama pull llama3.2 Open the Synaphos dashboard → Settings → set Backend URL to http://localhost:11434 → Test Connection → Save Go to Chat → start talking That is it. No account registration. No billing. No data leaving your machine. For remote servers, replace localhost with your server IP. The dashboard proxies all connections server-side, so HTTPS works with any HTTP backend.

Running Synaphos on a VPS what the production setup looks like Most AI tools assume you are on a local machine. Synaphos works equally well on a remote VPS. The setup is identical: install @ollama, pull a model, expose the dashboard over HTTPS, point it at your Ollama instance. On a VPS with 8 GB RAM running Ubuntu, llama3.2 (2 GB) loads in under 5 seconds and handles concurrent requests without issue. The first token appears in 1–2 seconds. The full proxy layer adds less than 20ms of overhead. We tested this on a $6/month VPS. No GPU required. Llama3.2 runs fine on CPU-only hardware with 8 GB RAM. If you want faster inference or larger models, scale up but the baseline works on commodity hardware that almost anyone can afford.

$SYPHOS Contract Audit Verified Safe on Base The $SYPHOS smart contract (0x5272653c6a50363E9300538497Ff8663CFAb4b07) has been independently audited by Serialized Audit and confirmed safe on Base mainnet. Token Audit: SAFE. Pool Audit: SAFE. Vulnerabilities found: 0. The contract was deployed through the CLANKER verified factory with no custom or unverified logic. Full report: serializedaudit.io/base/0x5272653… On the GMGN "Unsafe" label: GMGN flags tokens using automated heuristics token age, liquidity, and missing social metadata not contract-level security scanning. The $SYPHOS contract has no exploits, no honeypot, no hidden mint functions, and no admin keys. This is a metadata issue on GMGN's side, not a security finding. For verification, refer to the audit report directly.

Synaphos roadmap what is being built next Phase 1 is live: local AI chat, multi-model support via @ollama, EAS attestation on @BuildOnBase via @eas_eth, IPFS storage via @PinataCloud, wallet connect. Phase 2 introduces Sign-In With Ethereum (SIWE) for wallet-based authentication, SHA-256 output hashing for verifiable response integrity, and a self-hosted IPFS node option via @IpfSbot Kubo removing the @pinatacloud dependency entirely for users who want full sovereignty. Phase 3 brings the $SYPHOS governance layer on-chain: a custom ProvenanceRegistry contract on @BuildOnBase, community voting on protocol parameters, and the federated inference network where token holders can contribute compute and earn rewards. The goal is a protocol where AI provenance is not a product feature it is a public good governed by its users.

What the $SYPHOS token will do inside the app $SYNAPHOS is the native token of the Synaphos network. It is not a speculative asset it has a defined function inside the protocol from day one. Holding $SYPHOS grants governance rights over the Synaphos protocol: voting on schema updates to the on-chain attestation registry built on @eas_eth, whitelisted model providers, and future infrastructure decisions. Every attestation written on-chain is tied to the protocol's schema that schema is governed by token holders. In later phases, $SYPHOS will be required for certain on-chain actions: registering a custom attestation schema on @BuildOnBase, accessing federated inference from community-hosted nodes, and participating in the provenance marketplace where verified AI outputs can be indexed and discovered. The token connects usage to ownership the more you use the protocol, the more say you have in how it evolves.

Why we connect directly to Ollama, not a middleware layer Early versions of Synaphos routed through Odysseus as a middleware layer. In practice, this added latency and an unnecessary dependency for standard chat. @ollama already exposes a full OpenAI-compatible API at port 11434 /v1/models, /v1/chat/completions, streaming included. We now connect the dashboard directly to @ollama. Less moving parts, faster responses, and full compatibility with any OpenAI-compatible endpoint including @vllm_project , llama.cpp server, or @lmstudio The architecture is simple by design: browser → proxy → Ollama. The dashboard does not care what is running on the other end, as long as it speaks the OpenAI API format.

The proxy architecture that makes HTTPS + local AI work One problem nobody talks about: you cannot call a local HTTP server from an HTTPS page. The browser blocks it as a mixed content violation. This breaks every attempt to connect a hosted frontend to a local AI backend. We solved it with a server-side proxy. When the Synaphos dashboard is served over HTTPS and your @ollama instance runs over HTTP, all requests route through our API server first. The server makes the HTTP call internally and returns the response to the browser over HTTPS. No mixed content error. No browser restrictions. This means Synaphos works whether you are running @ollama on localhost, a home server, or a remote VPS without any browser extension, CORS workaround, or certificate setup on your end. $SYPHOS

How the on-chain layer works $SYPHOS Every AI response in Synaphos can be cryptographically attested on Base L2. When you click "Attest", two things happen: the full conversation is uploaded to IPFS via Pinata, and an EAS (Ethereum Attestation Service) record is written on-chain linking your wallet address, the model used, the IPFS CID, and a timestamp. The result is a permanent, tamper-proof record that proves: this output was generated by this model, at this time, by this address. Nobody can alter it. Nobody can delete it. The record lives on Base mainnet indefinitely. This is what AI provenance looks like in practice. Not a logo or a watermark an on-chain cryptographic attestation that anyone can verify independently.

What Synaphos actually is Synaphos is not another AI wrapper. It is a local-first workspace that runs entirely on your own hardware. You install Ollama, pull a model, point the dashboard at your machine, and start working. No account, no API key, no subscription. The entire stack runs on your device. Your conversations never touch a cloud server. The only time data leaves your machine is when you choose to attest a response on-chain and even then, you control exactly what gets published. We built it because every AI tool we tried either required handing data to a third party or was too complex to self-host. Synaphos is the middle ground: a clean UI for local AI that any developer can run in under 10 minutes. $SYPHOS

LOCAL AI IS NOT A COMPROMISE Running AI on your own hardware used to mean accepting worse results. That gap has closed. Models like Llama 3.2, Mistral, and Qwen 2.5 all running locally via Ollama are competitive with cloud APIs for the majority of everyday tasks. The difference is no longer capability. The difference is who controls the infrastructure. Synaphos runs any Ollama-compatible model with no rate limits, no per-token cost, and no account required. 8 GB of RAM is enough to get started. The model runs on your CPU or GPU. The response streams directly to your browser. Nothing else is involved.

OPEN SOURCE MEANS AUDITABLE "Trust us" is not a security model. When AI software is closed source, you are making a faith-based decision about what it does with your data. You cannot verify the claim. You have no recourse if the claim turns out to be false. Closed source and data privacy are structurally incompatible. Synaphos and Odysseus are fully open source under MIT license. Every dependency is declared. Every API call is visible in the code. The attestation logic, the IPFS upload path, the wallet integration all of it is readable, forkable, and verifiable. The code is the only honest answer to "can I trust this."

THE ATTESTATION IS YOURS When you click ATTEST in Synaphos, nothing happens automatically. Your wallet prompts you to sign. You approve. Only then does the conversation leave your machine once to IPFS for content storage, once to Base mainnet for the on-chain record. Two explicit actions. Both visible in your wallet history. The attestation is signed by your wallet address. It is listed on EASScan under your address. It links to your IPFS content. If Synaphos disappears tomorrow, the attestation remains readable by anyone, owned by no one except the address that created it.

We built Synaphos on top of Odysseus a self-hosted AI workspace with an on-chain layer. Every AI response can be attested on Base L2 via EAS and stored permanently on IPFS. No cloud. No API key. Runs entirely on your own hardware. Just posted our setup notes + a few doc issues we found in the wild to the Odysseus discussion board: github.com/pewdiepie-arch… Dashboard: synaphos.com Source: github.com/synaphos/synap…

@HanzoYasunaga we have sent you a message. Let's discuss.

@Synaphos Who is the dev send me dm

BUILD ON SYNAPHOS Synaphos is not a locked platform. The Odysseus backend exposes an OpenAI-compatible HTTP API at localhost:8000. Any tool that can talk to OpenAI can talk to Odysseus. You can point your existing scripts, IDE plugins, or automation pipelines at a local model with a one-line URL change. The full stack is MIT-licensed and documented. Fork it, embed it, strip out the parts you do not need, or build an entirely different interface on top of Odysseus. The codebase is TypeScript throughout with strict type checking. Contributions are open on GitHub at github.com/synaphos/synap….

@theOGrbspace We have opened DM for everyone to discuss

@Synaphos Yoooo yoooo, any team can dm me ? Im not a dumbshit marketer that offerin you a fake engage or useless services. Ive something big for you, i just bring some guy and they need to know more about what you guys build here. Shot me a dm, ngl it would be worth for us all

ABOUT THE PROJECT Synaphos started from a specific frustration: local AI tools had no provenance layer. You could run a model locally, but the output was indistinguishable from any other text. There was no way to prove what was generated, when, and by what model without trusting a third party. The EAS and IPFS stack was identified as the solution. EAS is an open, permissionless attestation standard. IPFS is content-addressed storage where the same CID always refers to the same content. Together they create a verifiable, permanent record that does not depend on any centralized service remaining online. Odysseus an existing open-source local AI backend was forked and adapted as the inference engine. The Synaphos dashboard was built on top: React, Vite, Tailwind, wagmi for wallet connection, and a minimal Express server to keep the Pinata JWT server-side. The full stack is TypeScript throughout. The project reached public launch in June 2026. The source is published on GitHub under MIT license. Phase 2 SIWE wallet authentication and community governance is on the roadmap and clearly labeled as not yet shipped.

WHY SYNAPHOS EXISTS Most AI tools are built around a business model that requires your data. The model needs to learn. The product needs to improve. The company needs to grow. You are not the customer you are the training set. This is not a conspiracy. It is a structural incentive built into the architecture of cloud AI products. Synaphos exists to offer a different structure. When you self-host, the incentive is gone. There is no server receiving your prompts. There is no company that benefits from your data. The software does its job and stays out of the way. The on-chain layer was added because trust needs to be verifiable, not assumed. If an AI-assisted output matters for research, for attribution, for legal documentation you should be able to prove its origin without relying on a company's logs. A cryptographic attestation on a public blockchain gives you that without requiring anyone's permission. Synaphos is MIT-licensed because the code being open is the only way to back up any of these claims. You do not have to take our word for it. You can read every line, verify every behavior, and deploy your own version if you disagree with any decision we make. That is what "open source" actually means not a marketing claim, but a structural commitment to accountability.

How to Self-Host AI on a VPS with Synaphos + Ollama (No Cloud, No API Key) In this video I walk through installing a fully self-hosted AI workspace on a VPS running Linux no paid API, no OpenAI account, no monthly subscription. Everything runs on your own hardware. What we cover: Installing Ollama on a VPS (Ubuntu/Debian) - Pulling a language model (llama3.2) - Connecting the Synaphos dashboard to the local Ollama instance - Testing the AI chat from a browser over HTTPS - Important you do NOT need a VPS to use this. - A VPS is just one option shown in this video. - Synaphos works on any machine that can run Ollama: your laptop, a home server, a Raspberry Pi, or any Linux/macOS/Windows machine. The setup is identical install Ollama, pull a model, point the dashboard at http://localhost:11434. Requirements: Any machine with 8 GB+ RAM Ollama installed and running A browser Links: Dashboard: synaphos.com/dashboard Setup guide: synaphos.com/dashboard (SETUP tab) Ollama: ollama.com Model library: ollama.com/library Source code: github.com/synaphos/synap… No data leaves your machine. No account needed. MIT licensed.

This is the official contract address for Synaphos the self-hosted, open-source AI workspace with native on-chain intelligence. $SYPHOS REAL CA : 0x5272653c6a50363E9300538497Ff8663CFAb4b07 clanker.world/clanker/0x5272…