Rate-limit model failover (switch mid-session to next in fallback chain, with cooldown) on both stream + invoke paths

Model selector (default minimax-m2.7, native tool-calling) + generous retries/patience (6 attempts, 180s timeout) (part 2)

Model selector (default minimax-m2.7, native tool-calling) + generous retries/patience (6 attempts, 180s timeout)

Raise depth caps (quick8/balanced12/deep20) so data queries complete despite redundant tool calls

Adaptations: web channel profile, depth control, markdown+skills+depth in Dexter, Ask-Dexter context passing, Research widgets board, RULES.md (part 2)

Adaptations: web channel profile, depth control, markdown+skills+depth in Dexter, Ask-Dexter context passing, Research widgets board, RULES.md

Fix dexter SSE disconnects: Bun idleTimeout 255s + proxy swallows mid-stream cutoffs (part 2)

Fix dexter SSE disconnects: Bun idleTimeout 255s + proxy swallows mid-stream cutoffs

Add flow zoom +/-/fit controls; harden proxy against transient upstream disconnects

LLM: freellmapi via openai provider (custom base URL), helmholtz stays Blablador

Fix: hedge board height (h-screen) + agent default model env override (part 2)

Fix: hedge board height (h-screen) + agent default model env override

Unified Dexter/HedgeFund/OpenBB app (FastAPI single-server, no Caddy/terminal) (part 3)

Unified Dexter/HedgeFund/OpenBB app (FastAPI single-server, no Caddy/terminal) (part 2)

Unified Dexter/HedgeFund/OpenBB app (FastAPI single-server, no Caddy/terminal)

initial commit