DevNet 1.0 — bugs & installation troubleshooting

[umermjd11]

Hit an error during install, setup, or day-to-day dincli use? Post it here — include the command you ran, the full error output, your OS/Python version.

This covers:

• Installation/setup issues (venv, wheel install, .env, wallet, IPFS provider)
• CLI errors during a Global Iteration (aggregator/auditor/client/model-owner flows)
• Anything that doesn't behave the way Documentation/setup.md or Documentation/GettingStarted.md says it should

Before posting, a quick search of open threads here saves everyone time — you may not be the first to hit it. Known, already-tracked issues (like the current slashAggregators GI-liveness bug described in GettingStarted.md) don't need a new thread unless you're hitting something additional.

See also: Setup Guide · Getting Started

[umeradl]

Flagging, not fixed yet: main's pyproject.toml is missing two dependencies that dincli imports directly — eth-account (dincli/cli/system.py:14, dincli/cli/utils.py:12, from eth_account import Account) and py-multibase (dincli/services/cid_utils.py:37, from multibase import encode). Neither is declared, even after the earlier dependency-declaration pass (#86) — that one caught rich/requests/click but missed these two.

This affects the dincli 0.2.0 wheel already built and pushed to origin/main — it inherits the same gap.

Currently low-severity: py-cid>=0.5.0 and web3>=6.0.0 both declare py-multibase/eth-account as their own hard requirements today, so a normal install still pulls them in transitively. But it's the same latent-dependency pattern that broke rich on develop (fixed in #97) — if py-cid or web3 ever demote either to optional, main breaks silently the same way.

Follow-up fix would just be adding py-multibase>=1.0.0 and eth-account>=0.13.0 to main's declared dependencies, mirroring #97's develop-side fix. Not done yet — flagging for visibility.

[umeradl]

Flagging, not fixed yet: the slashAggregators/minStake mismatch (GI-liveness bug — enforcement reverts because the coordinator's minStake is below the staking contract's MIN_STAKE floor, and since endGI requires the AggregatorsSlashed state slashAggregators can't produce, a missed/divergent aggregator submission wedges the Global Iteration entirely) is not code-fixed — only documented. Documentation/GettingStarted.md's Slashing section now says "a fix is planned soon," added this session.

The proposed real fix — redeploy platform-level contracts as upgradeable proxies, ship aggregator slashing disabled with an owner-only enable function, patch state on redeploy rather than resetting to zero — was posted to discussion #79. It's never been scoped into a PR.

Worth flagging: this was tracked as an explicit push blocker for main, and main has since been pushed to origin/main twice this session (the 0.2.0 version bump, and a separate setup.md fix) without the underlying contract fix landing — only the doc note satisfied the blocker at the time.