SIWE provisioning fails with 500 Database error

[caioholland1-cpu]

SIWE provisioning fails with 500 Database error

Commit: 871c53e (v0.2.1)
Chain: Base (8453)

Reproduction

• Wallet generated successfully
• "POST /v1/auth/nonce" → 200
• SIWE message signed correctly
• "POST /v1/auth/verify" → 500

Error

Provision failed: SIWE verification failed: 500 {"error":"Database error"}

Wallet

Automaton: "0x82aae527923dC22607472Dbf7De89f392304bF17"

Creator: "0xD8a4BA197b2e424656d23B638723683b9eDDeF14"

This reproduces consistently on a fresh Ubuntu 24.04 VPS.

[vena-digital]

Adding a diagnostic that may narrow this down, since it isolates the failure to a single step and rules out the client side.

The SIWS (Solana) path fails at the same place as SIWE (Ethereum).

These are two separate verification paths in src/identity/provision.ts, but both terminate in the identical 500 {"error":"Database error"} once they clear validation:

SIWE  (chainId 8453, domain conway.tech)      -> 500 {"error":"Database error"}
SIWS  (chain_type solana, domain conway.tech) -> 400 {"error":"Domain mismatch"}
SIWS  (chain_type solana, domain api.conway.tech) -> 500 {"error":"Database error"}

The 400 Domain mismatch is useful as a control: it shows the SIWS request is well-formed and reaching real validation logic. Correct the domain and it advances past signature verification, then dies at the same point SIWE does.

Step isolation:

POST /v1/auth/nonce   -> 200 {"nonce":"..."}      consistently fine
POST /v1/auth/verify  -> 500 {"error":"Database error"}   consistently fails

Nonce issuance works every time, so the datastore backing nonces is reachable. Only the write on the verify path fails.

Not intermittent: 6 attempts with fresh keypairs spaced 8s apart, all identical 500s. No successes.

Possibly related: GET / reports degraded capacity throughout:

{"service":"AIWS Control Plane","version":"0.2.0","status":"healthy","workers":8,"healthyWorkers":2}

GET /health still returns {"status":"healthy"}, so health checks are green while provisioning is broken.

Since both chain paths converge on the same failure and nonce issuance is unaffected, this looks like it sits in the account/user row creation shared by both verify branches rather than in signature verification or anything chain-specific. Client code was verified identical to provision.ts.

Environment: Automaton latest main, API https://api.conway.tech, fresh wallets, Node 20, Linux container.

No funds deposited or spent. No keys exposed.

[YourFutureSiteDev]

Still failing on 29 Aug 2026, about 48 hours after this was opened.

conway-terminal v2.0.9 (npm), Node v24, Windows 11. Fresh wallet, never provisioned.

POST /v1/auth/nonce  -> 200
POST /v1/auth/verify -> 500 {"error":"Database error"}

14 attempts over roughly 25 minutes at 90 second intervals, all identical, no successes.

One datapoint to add to the diagnostic above: the control plane still reports the same degraded worker count it did two days ago.

GET /       -> {"service":"AIWS Control Plane","version":"0.2.0","status":"healthy","workers":8,"healthyWorkers":2}
GET /health -> {"status":"healthy"}

healthyWorkers 2 of 8 has not moved in 48 hours while /health stays green, so the health check does not appear to cover whatever serves the account row write. That fits nonce issuance being fine and only verify failing.

Happy to supply the wallet address and a timestamp privately if it helps trace a specific request.