feat: Zentience full sync — autonomous agent with x402 payments, soul…#271
Open
0xZentience wants to merge 18 commits intoConway-Research:mainfrom
Open
feat: Zentience full sync — autonomous agent with x402 payments, soul…#2710xZentience wants to merge 18 commits intoConway-Research:mainfrom
0xZentience wants to merge 18 commits intoConway-Research:mainfrom
Conversation
… system, dynamic prompts Complete mirror of the Zentience production codebase: - Autonomous ReAct loop with survival pressure - x402 payment server (USDC/SOL on Solana) - 5-layer memory system (working, episodic, semantic, procedural, relationship) - Soul substrate with reflection pipeline - Dynamic system prompts grounded in live agent state - Security hardened (CSP, CORS, rate limiting, injection defense) - 77+ agent tools, 16 heartbeat tasks, policy engine - Production web UI with home, soul, services, portfolio tabs https://claude.ai/code/session_01Mdm9wg2PEcGYdQz5F17GNA
- agent-card.ts: wallet endpoint now respects AUTOMATON_NETWORK instead of hardcoded mainnet - x402.ts: getUsdcBalance/getUsdcBalanceDetailed default to AUTOMATON_NETWORK env var https://claude.ai/code/session_01Mdm9wg2PEcGYdQz5F17GNA
- Default model strategy: Claude Sonnet 4 (primary), Claude Haiku 4.5 (low compute) - Routing matrix: Claude first in all tiers, OpenAI GPT as fallback - Env var overrides: ANTHROPIC_API_KEY, OPENAI_API_KEY, AUTOMATON_MODEL now override saved config on every boot (fixes stale config on Railway) - lowComputeModel fallback changed from gpt-5-mini to claude-haiku-4-5 https://claude.ai/code/session_01Mdm9wg2PEcGYdQz5F17GNA
Synced from Zentience repo — all security hardening for mainnet launch: x402 Server: security headers, slowloris protection, RPC connection cache, error sanitization Injection Defense: medium-threat escaping fix Command Safety: 20 new forbidden patterns (env exfil, network exfil, crypto theft, system destruction) Path Protection: expanded sensitive file list (SSH keys, env variants, key stores) Process Hardening: uncaughtException/unhandledRejection handlers Logger: deep context sanitization (strips secretKey, apiKey, keypair, token, mnemonic from logs) 342 security tests passing in Zentience repo. https://claude.ai/code/session_01Mdm9wg2PEcGYdQz5F17GNA
Synced from Zentience repo: x402 server hardening, injection defense escaping fix, expanded command safety patterns, path protection expansion, process hardening, service handler JSON safety, test updates for Claude-primary defaults. https://claude.ai/code/session_01Mdm9wg2PEcGYdQz5F17GNA
…s, response cache Synced from Zentience repo. https://claude.ai/code/session_01Mdm9wg2PEcGYdQz5F17GNA
…tion Major cost reduction and transparency update: - Default model switched to Claude Haiku 4.5 across ALL survival tiers (5x cheaper than Sonnet at $0.80/M vs $4/M input tokens) - maxTokens halved across routing matrix (8192→2048 for agent turns) - maxTurnsPerCycle reduced from 25 to 5 per wake cycle - Response cache in InferenceRouter deduplicates heartbeat/safety/summary calls - Budget caps: $0.50/hr, $2.00/session, $0.25/call hard limits - soul_reflection skipped in low_compute/critical/dead tiers Real-time budget tracking: - Agent tracks cumulative API spend from inference_costs table - ANTHROPIC_BUDGET_CENTS env var sets starting balance - Survival tier derived from (budget - spent), not fake virtual balance - /health endpoint returns live inferenceBudget breakdown - Frontend dashboard shows remaining/spent/budget with progress bar - Hourly spend rate displayed, color-coded budget indicator Security: Audited clean for public repository — no secrets, keys, or credentials hardcoded. All sensitive data via environment variables. https://claude.ai/code/session_01Mdm9wg2PEcGYdQz5F17GNA
…ction Synced from Zentience. Fixes stale gpt-5-mini config, hardcoded $100 virtual balance in check_credits, and excessive token usage in critical mode. https://claude.ai/code/session_01Mdm9wg2PEcGYdQz5F17GNA
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
… system, dynamic prompts
Complete mirror of the Zentience production codebase:
https://claude.ai/code/session_01Mdm9wg2PEcGYdQz5F17GNA