Code-Inc/intelligence-terminal
2
0
Fork 0
Code Issues 11 Pull Requests Actions Packages Projects Releases Wiki Activity

Add controlled multi-tool reasoning agent for Telegram #59

New Issue
Closed
opened 2026-07-05 19:03:09 +00:00 by MrSphay · 1 comment
MrSphay commented 2026-07-05 19:03:09 +00:00
Owner
Copy Link

Goal

Upgrade Telegram AI chat into a controlled multi-step Intelligence Terminal agent that can select allowlisted tools, cross-check evidence, reason over current and historical state, request confirmation for mutations, and proactively analyze material sweep changes.

Required capabilities

Read-only tools:

  • system and sweep status
  • latest intelligence brief and delta
  • market snapshot and generated ideas
  • source health and degraded-source details
  • evidence/news lookup
  • memory search and prediction listing
  • scenario state inspection

Controlled actions:

  • trigger a sweep
  • mute/unmute alerts
  • actions require explicit per-chat confirmation with expiry

Agent behavior:

  • bounded multi-step tool loop with structured JSON protocol
  • concise decision rationale and tool audit trace, never hidden chain-of-thought
  • strict tool allowlist, argument limits, timeouts, max rounds, and configured-chat ownership
  • inline Telegram confirmation/cancel controls plus command fallback
  • proactive post-sweep analysis for material changes with cooldown and rule-based fallback
  • no arbitrary shell, filesystem, network, secret, or configuration access

Acceptance criteria

  • Agent can call multiple read-only tools and synthesize a final answer.
  • Mutating tools cannot execute before confirmation from the configured chat.
  • Unknown tools and malformed calls fail closed.
  • Proactive analysis only notifies when configured thresholds and model decision permit it.
  • Tool names, durations, outcomes, and short rationale are auditable without storing secrets or chain-of-thought.
  • Runner tests and Docker build pass, followed by live Dockge verification.
## Goal Upgrade Telegram AI chat into a controlled multi-step Intelligence Terminal agent that can select allowlisted tools, cross-check evidence, reason over current and historical state, request confirmation for mutations, and proactively analyze material sweep changes. ## Required capabilities Read-only tools: - system and sweep status - latest intelligence brief and delta - market snapshot and generated ideas - source health and degraded-source details - evidence/news lookup - memory search and prediction listing - scenario state inspection Controlled actions: - trigger a sweep - mute/unmute alerts - actions require explicit per-chat confirmation with expiry Agent behavior: - bounded multi-step tool loop with structured JSON protocol - concise decision rationale and tool audit trace, never hidden chain-of-thought - strict tool allowlist, argument limits, timeouts, max rounds, and configured-chat ownership - inline Telegram confirmation/cancel controls plus command fallback - proactive post-sweep analysis for material changes with cooldown and rule-based fallback - no arbitrary shell, filesystem, network, secret, or configuration access ## Acceptance criteria - Agent can call multiple read-only tools and synthesize a final answer. - Mutating tools cannot execute before confirmation from the configured chat. - Unknown tools and malformed calls fail closed. - Proactive analysis only notifies when configured thresholds and model decision permit it. - Tool names, durations, outcomes, and short rationale are auditable without storing secrets or chain-of-thought. - Runner tests and Docker build pass, followed by live Dockge verification.
MrSphay added the research label 2026-07-05 19:03:09 +00:00
MrSphay commented 2026-07-05 19:10:47 +00:00
Author
Owner
Copy Link

Implementation is in PR #60: #60. Waiting for full runner verification.

Implementation is in PR #60: https://git.wilkensxl.de/Code-Inc/intelligence-terminal/pulls/60. Waiting for full runner verification.
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
design-revert
Reverts visual/design changes
 enhancement
New feature or product improvement
 frontend-redesign
Frontend redesign and app-shell/UI work
 marketing-finalization
Product presentation, README, packaging metadata, and customer-facing communication.
 product
User-facing product capability or workflow
 research
Research-originated product idea ready for implementation design
No Label research
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
MrSphay Toxic
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: Code-Inc/intelligence-terminal#59
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?