ATT Infosoft — Internal PoC
Project Atlas
An AI agent system built to handle IT tender proposals end-to-end. Atlas orchestrates a fleet of specialist sub-agents, each owning one tender section. Together they produce competitive, review-ready submissions from a single RFP brief.
Orchestrator
Atlas
Production ready
Setup Phase
All phases complete
Skills Live
1 / 9
Hermes — TechSpec
GPU
RTX 5090
32GB VRAM — Local LLM
Agent Fleet Status
| Atlas | ● Production Ready — Orchestrator live on WhatsApp, memory persistent, tender workflow active |
| Hermes | ◑ Live — Quality Testing — First CAG run complete. Pending re-run on qwen3:32b to validate content depth before production use |
| Chronos | ◑ Built — Not Yet Run — SKILL.md complete (13 Mar). Covers PMP, Governance, Team Org + CVs. Next: first live tender run |
| Remaining (5) | Daedalus, Themis, Midas, Athena, Hephaestus, Hestia, Argus — Planned |
Objectives
Current focus
Validate Hermes output quality on qwen3:32b. Chronos (PMP) built — first live run next. Then Daedalus (Functional Response) to expand coverage.
Long-term vision
A full 9-agent fleet where Atlas reads an RFP and assembles a complete, submission-ready proposal — each section owned by a specialist agent, reviewed before send.
Hardware
| Processor | AMD Ryzen 9 9950X |
| Graphics Card | NVIDIA RTX 5090 32GB — Gigabyte |
| Memory | 64GB DDR5 6400MHz — KingBank RGB |
| Storage | 2TB Kioxia Gen.4 NVMe SSD |
| Motherboard | Asus TUF Gaming B650-E Wi-Fi 6E |
| Power Supply | Seasonic Vertex PX 1200W 80+ Platinum |
| Cooler | Corsair iCue Link Titan RX 360mm AIO |
| Operating System | Windows 11 Pro 64-bit |
| Network | Personal — off ATT infrastructure |
System Design
Architecture
Atlas runs a multi-agent architecture. The orchestrator reads the RFP, assigns tasks to sub-agents by skill, and assembles the final output. Model routing is cost-optimised across three tiers.
System Diagram
Agent Stack
Atlas — Main Agent (Orchestrator)
Powered by OpenAI Codex · GPT-5.4. Handles all user-facing chat, reads RFPs, assigns sections to sub-agents, and assembles the final tender output.
Codex Plus · $30/mo
Qwen3 32B — Primary Local Model
Running on RTX 5090 via Ollama. Q4_K_M quantization. Handles coding, analysis, document processing, tool calling, and tender drafting at zero compute cost.
Free · 24/7
Qwen2.5 Coder 32B — Code Specialist
Secondary local model for code generation, refactoring, debugging, and multi-language script automation. Activated by Atlas for code-heavy tasks.
Free · Local
Qwen3 8B — Fast Triage
Lightweight model for quick routing decisions and simple utility tasks. ~213 tok/s on RTX 5090. Minimal GPU load.
Free · Local
OpenRouter — Fallback Layer
Access to Claude, GPT-4o, and other models when a specific capability is needed beyond Codex or local LLM. Credit-based, used sparingly.
~$10–20 topup
Model Routing
| Task Type | Routed To | Model | Cost |
|---|---|---|---|
| Chat, coordination, RFP reading | Atlas (Cloud) | Codex · GPT-5.4 | $30/mo |
| Coding, analysis, doc processing | Local LLM (5090) | Qwen3 32B | Free |
| Code generation & refactoring | Local LLM (5090) | Qwen2.5 Coder 32B | Free |
| Quick triage & routing | Local LLM (5090) | Qwen3 8B | Free |
| Model-specific edge cases | OpenRouter | Claude / GPT-4o / others | ~$10–20 credit |
| Skill building, quality review | Claude Max (Gabriel, manual) | Claude Sonnet/Opus | Subscription |
Communication Channels
WhatsApp Group
Command centre. One group with Jackie, Sultan, Gabriel, and Atlas. All coordination, status updates, and decisions go through here. Managed via eSIM Bridge ($4.90/mo).
Individual DMs
Each person (Jackie, Sultan, Gabriel) has a separate DM with Atlas for detailed one-on-one work — deep task discussion, private reviews, and context-rich interactions without group noise.
MS Teams
Future plan only. Not included in current PoC — runs off ATT network on personal hardware. May be scoped in once PoC is validated with Jackie.
Memory Architecture
| Layer | Location | Used By | Notes |
|---|---|---|---|
| Shared memory | Monday.com private doc | Claude Chat, Cowork, Code, Atlas | Human-readable. Read at session start. |
| Local memory | MEMORY.md on 5090 | Atlas | Persistent state across Atlas sessions. |
| Agent rules | AGENTS.md on 5090 | Atlas | "Search memory before acting. Write decisions after every task." |
| Plugin | Supermemory (OpenClaw) | Atlas | Semantic memory retrieval within OpenClaw. |
| Decay | Temporal decay config | Atlas | Half-life: 30 days. Old memories fade over time. |
Feedback Loop
| Input | Past RFP and ATT submission uploaded as reference |
| Evaluate | Atlas compares submission vs RFP requirements, identifies gaps |
| Draft | Generates improved version based on evaluation |
| Review | PM reviews output and provides corrections |
| Store | Corrections saved to skill memory for next run |
Execution
Setup Plan
Four phases from hardware bring-up to first working skill. Currently in Phase 04.
01
Hardware Setup
- ✓Unbox and assemble PCAsus A32 case, Corsair AIO, RTX 5090, Kioxia 2TB SSD
- ✓First boot and BIOS checkEXPO enabled — RAM confirmed at 6400MHz. Kioxia 2TB detected. Mobo: Asus TUF B650-E BIOS 3827
- —Activate Windows 11 ProSkipped for now — no key bundled. Will activate when needed.
- ✓Run Windows Update fully
- ✓Install NVIDIA drivers (RTX 5090 — Blackwell)Driver 595.79 — installed via NVIDIA App
- ✓Verify GPU detected and CUDA enabledRTX 5090 confirmed in Device Manager
02
Core Stack Installation
- ✓Install Ollama v0.17.7 (Windows native)Windows native install, no WSL2
- ✓Pull Qwen3 32BPrimary local fallback — analysis, doc processing, tender drafting
- ✓Pull Qwen2.5 Coder 32BCode specialist — generation, refactoring, debugging
- ✓Pull Qwen3 8BFast triage fallback — ~213 tok/s on RTX 5090
- ✓Install OpenClaw 2026.3.11Windows native + Git for Windows dependency
- ✓OpenClaw onboardingGPT-5.4 primary, WhatsApp allowlist, Brave search API
- ✓Connect Codex OAuthCodex Plus — GPT-5.4
- ✓OpenClaw dashboardFixed — use `openclaw dashboard` to launch with auth
- ✓Configure model routingGPT-5.4 primary → Qwen3 8B → Qwen3 32B fallbacks
- ✓Install gateway serviceVia openclaw doctor --fix — auto-starts on boot
- —Connect OpenRouter API keyDeferred — not priority for now
- ✓Configure embedding providerOllama + nomic-embed-text (274MB) — free local embeddings
- ✓Add Jackie + Sultan to WA allowlistJackie (+6583229171), Sultan (+6593663700)
03
Atlas Identity and Channels
- ✓Create IDENTITY.mdName, role, emoji, stack, communication style per stakeholder
- ✓Create SOUL.mdCore behavioral rules, boundaries, proactive behavior, continuity
- ✓Create AGENTS.mdSession startup, memory rules, tender work rules, sub-agents, red lines
- ✓Configure MEMORY.md and memory/ directory on 5090Embedding: Ollama + nomic-embed-text, temporal decay 30-day half-life
- ✓Deregister WhatsApp from Mac MiniDone — number freed
- ✓Link WhatsApp number to Atlas on 5090+6598198569 linked via eSIM Bridge
- ✓Atlas first contact + tone calibrationAtlas alive on WA. Tone corrected from corporate to casual colleague. Preference saved to memory.
- ✓Memory persistence verifiedEnd-to-end: user speaks → Atlas adapts → writes memory file → git commits → persists
- ✓WA group policy configuredgroupPolicy: open, requireMention: false. SOUL.md controls listen-first reply behavior.
- ✓Profile picture generatedAI-generated portrait — team voted on final pick. Set as WA profile.
- ✓Create WhatsApp group"Project Atlas" — Jackie + Sultan + Gabriel + Atlas
- —Set up individual DMs with AtlasDeferred — not needed yet
04
Sub-Agent Fleet — Hermes First
- ✓Upload past tender RFPs as referencePA TACS, PMO CPIB, CAG T3 IIDS — in .openclaw/tender/
- ✓Upload past ATT submissions for format reference06 Technical Proposal + 07 System Design — in past-submission-reference/ per tender
- ✓Build Hermes TechSpec skill — SKILL.md, assets, evalsATT format: Book Antiqua, ATT logo header/footer, cover page, TOC. Evals: 2 iterations, 100% pass
- ✓Tender folder workflow — Atlas auto-discovers, no README neededOutputs auto-saved to tender submission/ subfolder
- ✓First live tender run — CAG T3 IIDS RefreshBoth docs produced (06 + 07), structure correct. Run used gpt-5.4 — model routing fix applied to AGENTS.md
- Re-run Hermes on qwen3:32b — validate content depthFirst run was thin (~32KB). qwen3:32b expected to produce richer, submission-ready output
- Review Hermes output with Jackie — sign off for production use
- ✓Build Chronos — PMP sub-agentSKILL.md, evals (3 test cases), reference docs, Schedule 6 CV template. Not yet run on live tender.
- First live Chronos run — CAG T3 IIDS PMP section
- Build next sub-agent — Daedalus (Functional Response)
Tender Skills
Skills Roadmap
One skill per tender section. Each skill is an independent module assigned to a sub-agent. Atlas activates skills based on the RFP. Skills improve through the feedback loop over time.
Skill Registry
Hermes — Technical Design
Generates the full technical design section: development approach, system architecture, cloud services setup, hosting environment, tech standards, and security design. Architecture is a sub-section here — not a separate document.
Daedalus — Functional Response
The largest section. Produces a feature-by-feature solution description mapped to every functional requirement in the RFP. Covers what the system does, how each requirement is met, UI workflows, and data handling. Typically 50–80 pages.
Themis — Compliance Matrix
Generates a paragraph-by-paragraph C / NC / PC compliance statement against every RFP clause — General, Functional, and ICT Security. Flags gaps and provides justifications for non-compliances.
Chronos — Project Management Plan
Generates project plan + Gantt, project governance & management, team organisation chart, roles/responsibilities, and CVs in Schedule 6 format. Covers CAG Sections 8–10 and PMO equivalent.
Midas — Pricing / BQ
Generates structured cost schedule based on scope, mandays, and rates. References past BQ formats and market benchmarks to produce competitive and defensible pricing.
Athena — Company Profile
ATT Infosoft background, past project references, and relevant track record. Tailored to the domain of each tender to maximise relevance score with evaluators.
Hephaestus — O&M + Training + Documentation
Generates the operations & maintenance plan (L1/L2 support structure, SLA compliance, patch management, hardware spares, VAPT, BCPs), training programme (schedule, syllabus, instructors), and documentation list. Confirmed as a consistent standalone section across PA, PMO, and CAG tender structures.
Hestia — Management Summary
High-level overview synthesised from all completed sections. Generated last. Frames ATT's value proposition and key differentiators for the evaluating committee.
Argus — Evaluator
Cross-checks every section against RFP requirements and the tender's evaluation criteria weightings. Scores submission quality, flags missing items, and highlights inconsistencies before final output.
Activity
Progress Log
Dated record of planning sessions, decisions, and milestones.
13 Mar 2026
Session 12 — Chronos built (PMP sub-agent)
Chronos SKILL.md built and registered. Covers CAG Sections 8–10: Project Plan (implementation schedule, SDLC phases, Gantt, QA plan, change management), Project Governance & Management (methodology, communication plan, risk register, contingency, exit plan), and Team Organisation & Competencies (org chart, roles/responsibilities, CVs in Schedule 6 format). Three eval cases written. Reference docs included: Schedule 6 CV template, standard risk register (10 risks), ICA AQS PMP structure. PMO CPIB past submission PMP not accessible (OneDrive sync issue) — built from CAG Appendix K requirements + ICA AQS real PMP. Ready for first live run on CAG tender.
13 Mar 2026
Session 11 — Hermes first live tender run (CAG T3 IIDS Refresh)
Hermes completed her first real tender run end-to-end. Atlas was pointed at .openclaw/tender/CAG, auto-discovered the requirements, spawned Hermes as a subagent, and produced both required documents — "06 Technical Proposal.docx" and "07 System Design and Technical Architecture.docx" — saved into tender submission/ subfolder. Structure confirmed correct: ATT logo in header, TOC, proper heading hierarchy, correct filenames. Content depth (gpt-5.4 run) is a baseline — next run will use qwen3:32b for richer output. Key learnings: model: frontmatter in SKILL.md is not auto-applied by OpenClaw; Atlas must explicitly pass the model in the subagent task prompt. Sub-agents are not visible in the Agents tab — use the Chat tab session dropdown to monitor a live subagent run. AGENTS.md updated with sub-agent model routing rules. Workspace backed up (108MB). Hermes milestone: sub-agent fleet operational.
13 Mar 2026
Session 10 — Hermes complete + gateway hardened
Hermes skill finalised: pinned to ollama/qwen3:32b (no API cost), ATT logo extracted from real CAG submission, professional docx output with Book Antiqua font / header / footer / cover page / TOC. Demo document generated and validated. CAG tender test folder set up at .openclaw/tender/CAG — Atlas auto-discovers documents and routes to correct agents with no README needed. Outputs saved to tender submission/ subfolder. Gateway incident: dropped at noon due to Scheduled Task having no restart policy. Atlas diagnosed and fixed it autonomously — new task XML with RestartCount=999, RestartInterval=PT1M, StartWhenAvailable=True. Gateway confirmed running post-fix.
13 Mar 2026
Session 8 — Phase 03 complete, Atlas operational
Atlas alive on WhatsApp — first contact successful. Tone calibrated from corporate to casual colleague (preference saved to Atlas memory, git committed). Memory persistence verified end-to-end. WA group policy configured: groupPolicy "open" + requireMention false — SOUL.md controls listen-first reply behavior. QMD memory attempted but incompatible with Windows (requires /bin/sh), reverted to builtin. Profile picture generated (AI portrait, team voted). WA group "Project Atlas" created with Jackie, Sultan, Gabriel, Atlas. Individual DMs deferred. Phase 03 complete.
13 Mar 2026
Session 7 — Atlas identity installed
Atlas persona files created and deployed to OpenClaw workspace: IDENTITY.md (name, role, stack, comm style), SOUL.md (behavioral rules, boundaries, proactive behavior), AGENTS.md (session startup, memory rules, sub-agents, red lines). Embedding provider configured: Ollama + nomic-embed-text for free local semantic search. WA allowlist updated with Jackie and Sultan. Phase 02 complete. Phase 03 in progress — remaining: WA number linking and group creation.
13 Mar 2026
Session 6 — Core stack operational
Dashboard issue resolved — gateway wasn't running; fixed with gateway.cmd and service install via openclaw doctor --fix. PowerShell execution policy set to RemoteSigned (was blocking CLI). All 3 Ollama models confirmed. Codex OAuth connected. Model routing configured: GPT-5.4 primary → Qwen3 8B → Qwen3 32B as fallbacks. GPT-5.4 kept as primary for stakeholder communication quality. OpenRouter deferred. Remaining: embedding provider for memory semantic search, WA allowlist for Jackie + Sultan.
12 Mar 2026
Session 2 — Architecture finalised
Communication strategy updated: WhatsApp group (Jackie + Sultan + Gabriel + Atlas) as command centre, with individual DMs per person. Discord dropped. Memory architecture decided: Monday.com private doc as shared memory across all Claude surfaces and Atlas. Local MEMORY.md + AGENTS.md + Supermemory plugin on 5090, with 30-day temporal decay. Sub-agents named (Greek mythology): Hermes, Midas, Argus, Themis, Daedalus, Chronos, Athena, Hestia. Multi-agent fleet (Titan, Forge) deferred — Atlas PoC first. GitHub (bcchew-art/Atlas) confirmed for code and site assets. Cowork confirmed able to deploy directly to GitHub.
12 Mar 2026
Session 1 — Project Atlas scoped
Defined full project scope, agent architecture, model routing strategy, and skill roadmap. Hardware confirmed as RTX 5090 workstation. Agent named Atlas. WhatsApp as primary channel. First milestone set as TechSpec skill. PoC to run off-network on personal hardware per Jackie's direction. Multi-agent fleet diagram designed.
People
Atlas Team
The orchestrator and her sub-agents. Each agent owns a specific tender skill and will be activated as skills are built.
Orchestrator
Atlas
Main Orchestrator
Active
ATT Infosoft's AI tender specialist. She analyses requirements, shapes proposal responses, and orchestrates specialist agents for tender development.
Sub-Agents
Hermes
TechSpec Writer
Live
ATT's technical design specialist. Give him an RFP and he'll translate it into a submission-ready architecture document — precise, structured, and built to score.
T
Themis
Compliance Matrix
Planned
D
Daedalus
Functional Response
Planned
M
Midas
Pricing / BQ
Planned
C
Chronos
Project Mgmt Plan
Built
A
Athena
Company Profile
Planned
H
Hephaestus
O&M + Training + Docs
Planned
H
Hestia
Executive Summary
Planned
A
Argus
Quality Evaluator
Planned
Platform
Known Issues
Running log of problems encountered, current workarounds, and notes for future OpenClaw fixes.
Open Issues
ISSUE-002
MEDIUM
Workaround in place
Sub-agent model routing ignored (model: frontmatter not applied)
OpenClaw does not honour the
model: field in SKILL.md when spawning a subagent. Hermes's first CAG run used gpt-5.4 (default) instead of ollama/qwen3:32b as intended. Workaround: Atlas must explicitly include the model in the subagent task prompt per the rule in AGENTS.md. Future fix: OpenClaw should apply SKILL.md model: frontmatter automatically at spawn time.
ISSUE-004
LOW
By design — documented
Sub-agents not visible in Agents tab
Sub-agents (Hermes etc.) are ephemeral runs, not registered agents — so they don't appear in the dashboard Agents tab. Workaround: use Chat tab → session dropdown → select the subagent session by label. Future fix: OpenClaw could show active subagent runs in a separate panel.
ISSUE-005
MEDIUM
Pending re-run
Hermes output thin on first run (gpt-5.4 baseline, ~32KB per doc)
First CAG run produced structurally correct docs but content depth was insufficient for real submission. Root cause: ran on gpt-5.4 due to ISSUE-002. Re-run with qwen3:32b expected to produce richer, submission-ready content. Also: SKILL.md could enforce minimum depth targets regardless of model.
Resolved Issues
ISSUE-001
HIGH
✓ Resolved 13 Mar 2026
Gateway outage goes undetected
Gateway dropped silently — no alert, no restart. Gabriel only noticed when WhatsApp messages got no reply. Root cause: Scheduled Task had no restart policy. Fix: Recreated task with RestartCount=999, RestartInterval=PT1M, StartWhenAvailable=True. Remaining gap: no proactive external alert if gateway is down for >5 minutes — morning recap is still the only passive health signal.
ISSUE-003
LOW
✓ Resolved 13 Mar 2026
Cannot self-update while gateway is running (EBUSY file lock)
npm threw EBUSY errors during openclaw update — gateway had files locked. Fix: documented update sequence in AGENTS.md — stop gateway first, back up, update, restart. Atlas prepares commands, Gabriel executes. Future fix: OpenClaw updater should handle graceful stop/update/restart as one atomic operation.