AI Agent 1.0.5 - Full Changelog
Released 2026-09-04 · v7.1.2.a · 3 issues · 11 features · 3 fixes
The complete record: every commit in the release, grouped by type, including work that never reaches a customer-facing note. The external changelog is the readable summary of the same release.
Jira References
-
DEV-2182 - AI API > Change Postgres DB default username and password
-
DEV-240 - AI API > Device does not support "NeighboringWiFiDiagnostic" > Ask AI chatbot > Run WiFi diagnostics > No information that diagnostic type not supported
-
DEV-2611 - Make sure to take care of various contradictions that appear in DeviceParams Tree
Features
-
Add demo logic to set X_ClientSteer to 1 upon successful steering execution
-
Normalize Google credentials path to absolute path relative to project root
-
Add discovery agent tool limit and configure interception and supervisor node environment variables
-
Add mesh network support and backhaul diagnostics to Wifi model and FTL client
-
Implement Wi-Fi issues diagnostic agent, playbook, and workflow integration with associated tests and prompt guidance
-
Add device parameter caching and integration to dynamic schema discovery
-
Implement isolated code execution sandbox for Python, Pandas, and Plotly analytics via FastAPI
-
Implement historical telemetry analytics agent with caching and expanded interceptor routing logic
-
Add sandbox service to docker compose, update build scripts, and documentation
-
Add support for fetching and processing historical device events via FTL API and sandbox runner
-
Add support for custom parameters in CPE models and improve agent streaming progress feedback
Bug Fixes
-
Add gittattributes
-
Enforce data integrity by prohibiting live tool substitution and auto-retries when historical telemetry is unavailable or empty
-
DEV-2611 - Make sure to take care of various contradictions that appear in DeviceParams Tree
Chores
-
Add pytest configuration and skip markers for resource-intensive integration and evaluation tests
-
Push ai-agent as :historical and the sandbox, emit digests
docker-push.sh now tags the main image ui/ai-agent:historical (the tag the env-manager sandbox lock keys on) and ui/ai-agent-sandbox:latest, and prints both pushed digests so the sandbox digest can be pinned in ft-charts-library. Made the builder setup idempotent and the script run from the repo root. (Was tagging :tmp.)
-
Add missing dependency pandas
-
Add docker-push-sandbox.sh
Refactoring
-
Centralize interrupt configuration and update stream handling to include localized preprocessing messages and improved tool execution status reporting
-
Disable tool restriction logic, update model temperatures, and enhance agent prompting for closed-loop verification
-
Simplify WiFi data retrieval by utilizing get_wifi_params and remove redundant host table processing
-
Mandate exclusive reporting of OEM metrics and update mock CPE data to include custom parameters
-
Improve type safety and add explicit type annotations across API clients, model logic, and agent workflows
Docker Images
ai-agent
- Image
-
hub.friendly-tech.com/ui/ai-agent:latest,v1.0.5-b0.0.1 - SHA256
-
sha256:3146eaf7276fc1daa1f32ce7ff2faa39e7c4f68fdb257c1aab8f325472c4f3e1
Dependencies
-
ACS → FT Configs Service
v1.0.2-b0.0.1 -
UI Backend → ACS
v7.1.2-b0.0.1 -
UI Portals → UI Backend
v7.1.2-b0.0.1 -
FT Device Network → ACS
v7.1.2-b0.0.1 -
UI AI Agent → UI Backend
v7.1.2-b0.0.1 -
UI AI Agent → FT Device Network
v1.0.3-b0.0.1