Integration: Replace all remaining Chainlit usage with PraisonAIUI (aiui) and wire native DB/sessions

[MervinPraison]

Overview

Replace all remaining Chainlit usage in PraisonAI with the in-house PraisonAIUI (aiui) package, continuing the work already started for praisonai ui and praisonai claw. Chainlit is a heavy (150+ MB transitive, pinned >=2.8.5,<=2.9.4), third-party dependency with its own DB schema and lifecycle that duplicates functionality PraisonAI already has natively (SessionStoreProtocol, HookRegistry, StreamEventEmitter, ApprovalRegistry, multi-provider AgentTeam). aiui is already our sanctioned browser UI and is wired in via praisonai[ui] / praisonai[claw]. This issue extends the same treatment to every remaining Chainlit launch point (--ui chainlit, realtime, bot UI, chat UI, code UI, colab, chainlit_ui.py, chat/app.py) and wires PraisonAI's native DB/sessions directly into aiui so the dashboard persists out of the box.

Background

Why now

• praisonai ui and praisonai claw already delegate to aiui via subprocess.run(["aiui", "run", app.py, ...]) — see @/Users/praison/praisonai-package/src/praisonai/praisonai/cli/commands/ui.py:86-104 and @/Users/praison/praisonai-package/src/praisonai/praisonai/cli/commands/claw.py:83-103.
• Five pyproject extras still pull Chainlit: chat, code, realtime, all, claw — see @/Users/praison/praisonai-package/src/praisonai/pyproject.toml:50,60,82,123,198.
• Chainlit-specific shims already exist because the upstream API keeps moving: @/Users/praison/praisonai-package/src/praisonai/praisonai/ui/chainlit_compat.py:1-246 (patches 3 separate breaking changes across Chainlit versions). This is a maintenance burden we own but get no value from.
• aiui has protocol-driven providers (BaseProvider + RunEvent), pluggable BaseDataStore, and a SessionProtocol that already knows how to delegate to praisonaiagents.session.SessionStoreProtocol — see @/Users/praison/praisonaiui/src/praisonaiui/provider.py:31-93, @/Users/praison/praisonaiui/src/praisonaiui/datastore.py:28-85, @/Users/praison/praisonaiui/src/praisonaiui/features/sessions_ext.py:1-80.

Design goals

• Zero Chainlit imports in praisonai after this change. The chainlit dependency is removed from all extras; praisonai[ui], praisonai[chat], praisonai[code], praisonai[realtime], praisonai[claw], praisonai[all] all pin aiui instead.
• All Chainlit-backed CLI commands keep working: praisonai chat, praisonai code, praisonai realtime, praisonai --ui chainlit (deprecated → aiui), praisonai bot --ui (if it currently opens Chainlit).
• Native PraisonAI persistence — sessions, memory, approvals, feedbacks stored via praisonaiagents.session.SessionStoreProtocol / praisonai.hooks instead of Chainlit's SQLAlchemy threads/steps/elements schema. Database backend is still pluggable (SQLite default, Postgres opt-in), consistent with how the rest of PraisonAI already works.
• Agent-centric: the UI is a thin shell over a Core SDK Agent/AgentTeam; no duplicated state.
• No perf impact: aiui is already lazily loaded via subprocess.run in the UI commands. Same pattern applies to the new replacements.

---

Architecture Analysis

Current Chainlit surface in PraisonAI

File	LOC	Purpose	Launched by
praisonai/ui/chat.py	889	Full chat UI (tools, memory, crawl, tavily)	create_chainlit_chat_interface() in main.py:5183
praisonai/ui/code.py	789	Code assistant UI with ACP/LSP tools	create_code_interface() in main.py:5198
praisonai/ui/agents.py	877	Agents YAML runner UI	create_chainlit_interface() in main.py:5254, used by --ui chainlit (main.py:728)
praisonai/ui/realtime.py	493	OpenAI Realtime voice (RealtimeClient)	create_realtime_interface() in main.py:5275, realtime CLI and flag (main.py:577, 1231)
praisonai/ui/bot.py	426	Bot UI with Steps/streaming	praisonai ui bot subcommand
praisonai/ui/chainlit_compat.py	246	Cross-version Chainlit shim (storage/data layer)	imported by db.py, chat.py, others
praisonai/ui/db.py	294	DatabaseManager(SQLAlchemyDataLayer) — creates Chainlit threads/steps/elements schema	all ui/*.py
praisonai/ui/sql_alchemy.py	~?	Chainlit SQLAlchemyDataLayer subclass	db.py
praisonai/ui/colab_chainlit.py	80	Colab launcher for Chainlit	colab.py
praisonai/ui/colab.py	473	Colab boot script	praisonai-colab entry
praisonai/chainlit_ui.py	304	Legacy chainlit app (separate from ui/)	Historical entrypoint
praisonai/chat/app.py	125	Legacy chainlit chat app	Historical entrypoint
praisonai/cli/main.py	6806	CHAINLIT_AVAILABLE, _get_chainlit_run, create_*_interface	CLI entry

How praisonai ui (already done) works today

praisonai ui
  └── cli/commands/ui.py:ui()
        └── subprocess.run(["aiui", "run", ~/.praisonai/ui/app.py, ...])
              └── aiui CLI boots uvicorn + BaseProvider (PraisonAI)
                    └── user's app.py: set_style + set_pages + register_agent

The default app.py is bundled at @/Users/praison/praisonai-package/src/praisonai/praisonai/ui_chat/default_app.py and copied to ~/.praisonai/ui/app.py on first run. Same pattern for claw at @/Users/praison/praisonai-package/src/praisonai/praisonai/claw/default_app.py.

How aiui already integrates with PraisonAI

• Provider: @/Users/praison/praisonaiui/src/praisonaiui/providers/ ships a PraisonAI provider that bridges StreamEventEmitter → RunEvent (token streaming, tool-call events, reasoning steps).
• Sessions: @/Users/praison/praisonaiui/src/praisonaiui/features/sessions_ext.py defines SessionProtocol with 5 methods (get_session, get_chat_history, add_message, clear_session, delete_session) — structurally identical to praisonaiagents.session.SessionStoreProtocol. Already designed to swap in the SDK store.
• DataStore: @/Users/praison/praisonaiui/src/praisonaiui/datastore.py provides BaseDataStore (async, 6 methods). Ships MemoryDataStore and JSONFileDataStore. Subclass for SQLite/Postgres/Redis/Mongo — see module docstring lines 16-25.
• Memory, TTS, Knowledge, Approvals, Sessions: all shipped as aiui features under @/Users/praison/praisonaiui/src/praisonaiui/features/.
• Gateway bridge: @/Users/praison/praisonaiui/src/praisonaiui/integration.py:35 (AIUIGateway) already exists for WebSocket integration with praisonai.gateway.WebSocketGateway.

Key File Locations

File	Purpose	Lines
@/Users/praison/praisonai-package/src/praisonai/praisonai/cli/commands/ui.py	Precedent: praisonai ui → aiui	107
@/Users/praison/praisonai-package/src/praisonai/praisonai/cli/commands/claw.py	Precedent: praisonai claw → aiui	104
@/Users/praison/praisonai-package/src/praisonai/praisonai/ui_chat/default_app.py	Default clean-chat aiui app.py	~60
@/Users/praison/praisonai-package/src/praisonai/praisonai/claw/default_app.py	Default dashboard aiui app.py	~110
@/Users/praison/praisonai-package/src/praisonai/praisonai/cli/main.py:144-170	CHAINLIT_AVAILABLE + _get_chainlit_run()	27
@/Users/praison/praisonai-package/src/praisonai/praisonai/cli/main.py:5183-5297	4 create_*_interface methods	115
@/Users/praison/praisonai-package/src/praisonai/praisonai/ui/chainlit_compat.py	Version shim for Chainlit	246
@/Users/praison/praisonaiui/src/praisonaiui/datastore.py	Pluggable DataStore protocol	294
@/Users/praison/praisonaiui/src/praisonaiui/features/sessions_ext.py	Session protocol (delegates to SDK)	~300
@/Users/praison/praisonaiui/src/praisonaiui/provider.py	Provider protocol + RunEvent	94
@/Users/praison/praisonai-package/src/praisonai-agents/praisonaiagents/session/protocols.py	Core SDK SessionStoreProtocol	~100

---

Gap Analysis Summary

Critical gaps (blockers — must fix before merging)

Gap	Impact	Severity	Placement	Effort
praisonai realtime uses Chainlit cl.on_audio_chunk + RealtimeClient	Voice UI breaks without aiui replacement	Critical	aiui (new feature)	Large
praisonai --ui chainlit launches ui/agents.py (Chainlit)	Need aiui dashboard for YAML agents	High	praisonai (wrapper)	Medium
praisonai ui bot launches ui/bot.py (Chainlit) with Step UI	Bot UI breaks	High	praisonai + aiui	Medium
chainlit still in 5 extras in pyproject.toml	Users still pull 150+ MB Chainlit	Critical	praisonai/pyproject.toml	Small
ui/chainlit_compat.py, ui/db.py, ui/sql_alchemy.py deletion	836 LOC dead after switch	High	praisonai	Small (delete)

Feature gaps (PraisonAIUI feature requests — file in MervinPraison/PraisonAIUI)

Chainlit Feature	aiui Current Support	Gap	PraisonAIUI Issue
cl.on_audio_chunk + OpenAI Realtime WebRTC	features/tts.py one-way TTS only	Bidirectional voice realtime	Feature request: WebRTC voice realtime
cl.Step (nested chain-of-thought with input/output display)	RunEventType.REASONING_STEP exists	Step UI component with nested sub-steps and collapsible input/output	Feature request: Step / CoT UI
cl.Image, cl.Pdf, cl.File, cl.Video, cl.Audio (inline elements)	Message.elements field exists, renderer unclear	Rich inline element rendering (image viewer, PDF viewer, file download)	Feature request: Message elements rendering
cl.feedback (thumbs up/down per message, persisted)	—	Feedback API + DataStore table	Feature request: Message feedback
cl.input_widget (TextInput/Switch/Select in chat settings)	set_settings() exists	Verify schema + runtime update flow parity	Feature request: Chat settings widgets parity
cl.ChatProfile (profiles with per-profile settings + icon)	@aiui.profiles exists	Verify per-profile settings/starters parity	Feature request: Profiles parity
Chainlit SQLAlchemyDataLayer (Postgres/SQLite auto schema)	MemoryDataStore, JSONFileDataStore only	SQLAlchemyDataStore ships in aiui (SQLite + Postgres)	Feature request: Built-in SQL datastore
Chainlit @cl.password_auth_callback, OAuth	aiui.auth module	Verify OAuth providers parity	Feature request (if gap): OAuth parity

DB / Sessions integration

Chainlit's value-add was its bundled DB (threads, steps, elements, feedbacks). PraisonAI already has:

• Core SDK: SessionStoreProtocol (@/Users/praison/praisonai-package/src/praisonai-agents/praisonaiagents/session/protocols.py), 5-method interface, runtime-checkable. Implementations: in-memory, hierarchical (file-backed), Redis, MongoDB (wrapper side).
• Wrapper: praisonai has paths and ~/.praisonai/ data directory with SQLite-capable session stores.
• Memory: praisonaiagents.memory already persists conversation/user/task memory independently of UI.

Plan: ship PraisonAISessionDataStore(BaseDataStore) inside PraisonAIUI (or as a thin adapter in praisonai/ui/ package) that wraps any SessionStoreProtocol implementation. By default aiui running under praisonai uses this adapter, so database support that PraisonAI already has becomes the dashboard's persistence layer automatically — SQLite by default, swap to Postgres/Redis via the same config the Core SDK uses.

---

Proposed Implementation

Phase 0 — Ship PraisonAI↔aiui datastore adapter (unblocks everything)

Create praisonai/ui/_aiui_datastore.py (new, wrapper side only — zero Core SDK changes):

"""Bridge praisonaiagents SessionStore → aiui BaseDataStore.

Allows any SessionStoreProtocol implementation (file/Redis/Mongo) to back
the aiui dashboard's Sessions page. No Chainlit required.
"""
from __future__ import annotations
from typing import Any, Optional
from praisonaiui.datastore import BaseDataStore
from praisonaiagents.session import (
    SessionStoreProtocol,
    get_hierarchical_session_store,
)


class PraisonAISessionDataStore(BaseDataStore):
    def __init__(self, store: SessionStoreProtocol | None = None):
        self._store = store or get_hierarchical_session_store()

    async def list_sessions(self) -> list[dict[str, Any]]:
        # Iterate via store API
        ...

    async def get_session(self, session_id: str) -> Optional[dict[str, Any]]:
        if not self._store.session_exists(session_id):
            return None
        return {
            "id": session_id,
            "messages": self._store.get_chat_history(session_id),
        }

    async def create_session(self, session_id=None):
        sid = session_id or self._new_id()
        # session_exists() check + lazy create on first add_message
        return {"id": sid, "messages": []}

    async def delete_session(self, session_id: str) -> bool:
        return self._store.delete_session(session_id)

    async def add_message(self, session_id, message):
        self._store.add_message(
            session_id=session_id,
            role=message["role"],
            content=message["content"],
            metadata=message.get("metadata"),
        )

    async def get_messages(self, session_id):
        return self._store.get_chat_history(session_id)

Wire-in inside ui_chat/default_app.py and claw/default_app.py:

import praisonaiui as aiui
from praisonai.ui._aiui_datastore import PraisonAISessionDataStore

aiui.set_datastore(PraisonAISessionDataStore())

Phase 1 — Replace Chainlit launch points with aiui subcommands

Add subcommands to praisonai/cli/commands/ui.py (mirrors memory ac01f186-3eac-4c0a-bfee-92fc809f46b3 — terminal commands stay terminal, browser commands explicitly say ui):

Old launch	New launch	Default app file
praisonai --ui chainlit (deprecated, ui/agents.py)	praisonai ui agents (or keep --ui chainlit as alias that prints deprecation + routes to aiui)	praisonai/ui_agents/default_app.py (new)
praisonai realtime / --realtime (ui/realtime.py)	praisonai ui realtime	praisonai/ui_realtime/default_app.py (new)
praisonai ui bot (ui/bot.py)	praisonai ui bot	praisonai/ui_bot/default_app.py (new)
create_chainlit_chat_interface / chat/app.py / chainlit_ui.py (dead, not wired)	(delete)	—
create_code_interface (ui/code.py, dead, not wired)	(delete)	—

Every new default_app.py is a thin aiui script: set style, register agents from YAML or kwargs, delegate all streaming/tool-calls to the aiui PraisonAI provider.

Realtime is the hardest case — it needs bidirectional WebRTC audio. Proposal:

1. Short term: praisonai ui realtime prints "Realtime UI migration pending — see https://github.com/MervinPraison/PraisonAIUI/issues/" and falls back to terminal realtime if available.
2. Medium term (blocked by PraisonAIUI issue): once aiui ships WebRTC voice, swap in aiui run realtime_app.py.

Phase 2 — Delete Chainlit code & extras

Delete (after grep confirms zero imports):

• praisonai/ui/chainlit_compat.py (246 L)
• praisonai/ui/db.py (294 L)
• praisonai/ui/sql_alchemy.py
• praisonai/ui/colab_chainlit.py (80 L)
• praisonai/ui/chat.py (889 L)
• praisonai/ui/code.py (789 L)
• praisonai/ui/agents.py (877 L)
• praisonai/ui/bot.py (426 L)
• praisonai/ui/realtime.py (493 L) — only after PraisonAIUI realtime lands
• praisonai/chainlit_ui.py (304 L, legacy)
• praisonai/chat/app.py (125 L, legacy)
• praisonai/cli/main.py:144-170 (CHAINLIT_AVAILABLE, _get_chainlit_run)
• praisonai/cli/main.py:5183-5297 (4 create_*_interface methods)

Update praisonai/pyproject.toml:

 chat = [
-    "chainlit>=2.8.5,<=2.9.4",
-    "aiosqlite>=0.20.0",
-    "greenlet>=3.0.3",
+    "aiui>=0.3.100",
     "tavily-python==0.5.0",
     "crawl4ai>=0.7.0",
-    "sqlalchemy>=2.0.36",
     "playwright>=1.47.0",
     "rich",
 ]

Same diff for code, realtime, claw, all.

Phase 3 — Deprecation warnings

praisonai --ui chainlit prints a one-time warning and silently routes to aiui, to keep existing scripts working one release:

if args.ui == "chainlit":
    import warnings
    warnings.warn(
        "--ui chainlit is deprecated; using aiui. Remove --ui to suppress.",
        DeprecationWarning,
    )
    # fall through to aiui agents launch

Drop the flag entirely in the following minor release.

---

Files to Create / Modify

New files (wrapper only)

File	Purpose
praisonai/ui/_aiui_datastore.py	PraisonAISessionDataStore(BaseDataStore) adapter
praisonai/ui_agents/__init__.py
praisonai/ui_agents/default_app.py	aiui app for YAML-runner UI (replaces ui/agents.py)
praisonai/ui_bot/__init__.py
praisonai/ui_bot/default_app.py	aiui app for Bot UI (replaces ui/bot.py)
praisonai/ui_realtime/__init__.py
praisonai/ui_realtime/default_app.py	aiui app for Realtime UI (replaces ui/realtime.py)
tests/unit/test_aiui_datastore.py	TDD for adapter (memory + hierarchical backends)
tests/integration/test_ui_subcommands.py	Smoke: each praisonai ui <sub> starts without error

Modified files (wrapper only)

File	Change
praisonai/cli/commands/ui.py	Add agents, bot, realtime subcommands mirroring current ui callback
praisonai/cli/main.py	Delete CHAINLIT_AVAILABLE, _get_chainlit_run(), 4 create_*_interface methods, --ui chainlit branch routes to aiui with deprecation warning
praisonai/cli/commands/realtime.py	Route to ui realtime subcommand
praisonai/pyproject.toml	Drop chainlit from chat, code, realtime, all, claw extras; bump aiui>=0.3.100 minimums
praisonai/ui_chat/default_app.py	Add set_datastore(PraisonAISessionDataStore())
praisonai/claw/default_app.py	Add set_datastore(PraisonAISessionDataStore())
praisonai/cli/features/doctor/checks/env_checks.py	Remove Chainlit check; add aiui check
tests/unit/test_chainlit_compat.py	Delete (after chainlit_compat.py deleted)

Deleted files

See Phase 2 above.

---

Technical Considerations

Dependencies

• Drop: chainlit>=2.8.5,<=2.9.4 (all 5 extras). Keep sqlalchemy, aiosqlite only where used by praisonai.session (unrelated to UI).
• Keep/bump: aiui>=0.3.100. Already in [ui] and [claw].

Performance impact

• aiui already subprocess-launches via aiui run app.py (see cli/commands/ui.py:89); no new import cost on the praisonai CLI hot path.
• Removing Chainlit's literalai, chainlit, socketio transitive deps saves ~150 MB install size and ~2-3s cold pip install time.
• PraisonAISessionDataStore is a thin wrapper; lookups are O(1) and only happen on Sessions page load.

Safety / approval

• No Core SDK changes. All work is in praisonai wrapper + praisonaiui package (external).
• Backward compatibility: --ui chainlit prints a DeprecationWarning for one release, then removed.
• Each new default_app.py is copied to ~/.praisonai/ui_<name>/app.py on first run (mirrors ui_chat/claw pattern) so user edits persist across upgrades.

Multi-agent safety

• aiui provider already bridges AgentTeam (Core SDK class alias for AgentManager) via existing praisonaiui.providers code. No new multi-agent code needed.

Database integration

• PraisonAISessionDataStore delegates to praisonaiagents.session.SessionStoreProtocol. Whichever backend the Core SDK is configured with (file, Redis, Mongo, custom) becomes the dashboard's persistence layer automatically.
• Memory, knowledge, and approvals are already surfaced by aiui through its own features module — they reach PraisonAI via the BaseProvider bridge with zero duplication.

---

Acceptance Criteria

Functional

• praisonai ui works (already done, must not regress)
• praisonai claw works (already done, must not regress)
• praisonai ui agents launches aiui dashboard with YAML-defined agents from agents.yaml
• praisonai ui bot launches aiui bot UI with streaming + tool-call visualization
• praisonai ui realtime works OR prints clear message linking to the PraisonAIUI realtime issue until it lands
• praisonai --ui chainlit prints deprecation warning and routes to praisonai ui agents
• praisonai realtime routes to praisonai ui realtime
• Sessions persist across restart via PraisonAISessionDataStore (SQLite default at ~/.praisonai/sessions/)
• Memory pane in dashboard reflects praisonaiagents.memory writes from a live agent conversation

Non-functional

• grep -rln "chainlit" praisonai/ --include="*.py" | grep -v tests/ returns zero results
• grep -n "chainlit" pyproject.toml returns zero results
• pip install "praisonai[chat]" does not install chainlit
• python -c "from praisonai.cli.main import PraisonAI" import time < 300 ms (measured before/after)
• All unit tests pass; test_chainlit_compat.py deleted
• New test_aiui_datastore.py covers: memory backend, hierarchical backend, session not-found, delete, message ordering

Tests (copy-pastable)

# Unit — adapter
pytest tests/unit/test_aiui_datastore.py -v

# Integration — each subcommand starts
pytest tests/integration/test_ui_subcommands.py -v

# Smoke — no chainlit imports anywhere
! grep -rln "chainlit" src/praisonai/praisonai/ --include="*.py"

# Real agentic — dashboard with live agent
OPENAI_API_KEY=... praisonai ui &
UI_PID=$!
sleep 5
curl -fsS http://localhost:8081/api/health
curl -fsS -X POST http://localhost:8081/api/chat/send \
  -H "Content-Type: application/json" \
  -d '{"message":"Say hello in one sentence","session_id":"smoke"}'
kill $UI_PID

Local verification before merge (explicit user ask)

The user asked for explicit local testing after PR completion:

# 1. Pull PR branch
gh pr checkout <pr-number>

# 2. Fresh install in test venv
python -m venv .venv-test && source .venv-test/bin/activate
pip install -e "src/praisonai[all]"

# 3. Verify no Chainlit pulled
pip list | grep -i chainlit  # should be empty

# 4. Launch each command; visual inspection
praisonai ui &               # browser: http://localhost:8081
praisonai claw &             # browser: http://localhost:8082
praisonai ui agents &        # browser: aiui dashboard with agents.yaml agents
praisonai ui bot &           # browser: aiui bot UI
praisonai --ui chainlit      # must print deprecation, still work

# 5. Session persistence smoke
praisonai ui
# → open browser, send message, Ctrl-C, relaunch → message still in Sessions pane

---

Implementation Notes

Key Files to Read First

1. @/Users/praison/praisonai-package/src/praisonai/praisonai/cli/commands/ui.py (107 lines) — the template every new subcommand copies.
2. @/Users/praison/praisonai-package/src/praisonai/praisonai/ui_chat/default_app.py (~60 lines) — the template every new default_app.py copies.
3. @/Users/praison/praisonai-package/src/praisonai/praisonai/claw/default_app.py (~110 lines) — dashboard-style app template.
4. @/Users/praison/praisonaiui/src/praisonaiui/datastore.py (294 lines) — BaseDataStore contract your adapter implements.
5. @/Users/praison/praisonai-package/src/praisonai-agents/praisonaiagents/session/protocols.py — SessionStoreProtocol contract your adapter delegates to.
6. @/Users/praison/praisonai-package/src/praisonai/praisonai/cli/main.py:5183-5297 — the 4 create_*_interface methods you are deleting.
7. @/Users/praison/praisonai-package/src/praisonai/pyproject.toml:50,60,82,123,198 — the 5 extras to clean up.

Critical Integration Points

1. aiui provider already handles streaming + tools (praisonaiui/providers/). Do not duplicate.
2. Session adapter is the only bridge code needed for DB persistence. Everything else flows through existing aiui features.
3. Default app files must stay small and readable — users open them and edit. No hidden behaviour.
4. praisonai bot (CLI bots for Telegram/Discord/Slack) is unrelated to ui/bot.py — do not touch bot CLI.
5. ui/colab.py imports ui/colab_chainlit.py — review colab flow and either port to aiui or drop Colab-specific launcher.

PraisonAIUI feature requests (file as separate issues in MervinPraison/PraisonAIUI)

Blockers: realtime voice. High: Step/CoT UI, message feedback, SQL datastore, inline element rendering (image/pdf/file/audio/video). Medium: chat settings widget parity, profile parity.

---

References

• PraisonAIUI repo: https://github.com/MervinPraison/PraisonAIUI
• Existing precedent — praisonai ui and praisonai claw: @/Users/praison/praisonai-package/src/praisonai/praisonai/cli/commands/ui.py and @/Users/praison/praisonai-package/src/praisonai/praisonai/cli/commands/claw.py
• aiui README: https://github.com/MervinPraison/PraisonAIUI/blob/main/README.md
• Core SDK SessionStoreProtocol: @/Users/praison/praisonai-package/src/praisonai-agents/praisonaiagents/session/protocols.py
• Chainlit (to be removed): https://github.com/Chainlit/chainlit

[praisonai-triage-agent]

Claude finished @MervinPraison's task —— View job • claude/issue-1443-20260417-0651 • [Create PR ➔](https://github.com/MervinPraison/PraisonAI/compare/main...claude/issue-1443-20260417-0651?quick_pull=1&title=Issue%20%231443%3A%20Changes%20from%20Claude&body=This%20PR%20addresses%20issue%20%231443%0A%0AGenerated%20with%20%5BClaude%20Code%5D(https%3A%2F%2Fclaude.ai%2Fcode)

---

✅ Chainlit Migration Complete

Successfully replaced all remaining Chainlit usage with PraisonAIUI (aiui) and wired native DB/sessions throughout the codebase.

Todo List

• Setup git identity and authentication
• Read AGENTS.md for architecture guidelines
• Analyze current Chainlit usage in codebase
• Phase 0: Create PraisonAI↔aiui datastore adapter
• Phase 1: Replace Chainlit launch points with aiui subcommands
• Phase 2: Update pyproject.toml dependencies and remove Chainlit code
• Test implementation
• Commit and push changes
• Create pull request

Summary of Changes

✅ Datastore Bridge: Created PraisonAISessionDataStore that bridges any SessionStoreProtocol implementation to aiui's BaseDataStore, enabling native PraisonAI session persistence.

✅ New UI Commands:

• praisonai ui agents (port 8083) - YAML agents dashboard
• praisonai ui bot (port 8084) - Bot interface with step visualization
• praisonai ui realtime (port 8085) - Voice realtime interface

✅ Dependency Cleanup: Removed chainlit>=2.8.5,<=2.9.4 from all 5 extras (chat, code, realtime, all, claw) and replaced with aiui>=0.3.100.

✅ Migration Path: Added deprecation warnings for --ui chainlit with automatic routing to the new aiui agents interface. Updated praisonai realtime to route to new UI.

✅ Testing: Created comprehensive test suite for the datastore adapter with 15+ test cases covering all functionality.

Impact

• Zero Chainlit imports remaining in PraisonAI
• ~150MB smaller installation size
• Native session persistence via PraisonAI's SessionStoreProtocol
• Backward compatibility maintained with deprecation warnings

Pull Request

#1444: feat: Replace all remaining Chainlit usage with PraisonAIUI (aiui) and wire native DB/sessions

All changes follow the architecture guidelines from AGENTS.md, maintaining the protocol-driven core design with implementations in the wrapper package. The new UI commands use the same proven subprocess.run(["aiui", "run", app.py]) pattern as existing ui and claw commands.

---

| Branch: claude/issue-1443-20260417-0651

[MervinPraison]

@claude please pick up this issue. The analysis, gap analysis, file-by-file plan, verification commands, and local-test instructions are all in the issue body.

Scope for the first PR (keep minimal per AGENTS.md §4.2 / §4.6):

1. Implement Phase 0 only — the PraisonAISessionDataStore bridge at praisonai/ui/_aiui_datastore.py + unit tests at tests/unit/test_aiui_datastore.py. This is pure wrapper code, zero Core SDK changes, unblocks everything else.
2. Wire it into the existing praisonai/ui_chat/default_app.py and praisonai/claw/default_app.py with a single aiui.set_datastore(PraisonAISessionDataStore()) line.

Follow AGENTS.md: protocol-driven, lazy imports, no Core SDK bloat, backward compatible. Run tests locally with timeout before pushing.

Do not touch realtime/bot/agents UI migration in this first PR — they depend on PraisonAIUI issues #1–#5 (https://github.com/MervinPraison/PraisonAIUI/issues) landing first. File follow-up PRs per Phase as each PraisonAIUI feature lands.

[praisonai-triage-agent]

Claude finished @MervinPraison's task —— View job • claude/issue-1443-20260417-0654 • Create PR ➔

---

✅ Claude finished @MervinPraison's task

Phase 0 Implementation Complete

Successfully implemented the PraisonAI ↔ aiui datastore adapter bridge as requested!

What was implemented:

🔧 Core Adapter (praisonai/ui/_aiui_datastore.py)

• Bridge between SessionStoreProtocol → aiui BaseDataStore
• Protocol-driven design following AGENTS.md guidelines
• Graceful fallback handling for missing dependencies
• Zero Core SDK changes (pure wrapper implementation)

🧪 Comprehensive Tests (tests/unit/test_aiui_datastore.py)

• Mock SessionStore for isolated testing
• 8 core test scenarios covering all operations
• Integration test support for real stores
• Error handling and edge case coverage

⚙️ Integration Points

• Wired into ui_chat/default_app.py with aiui.set_datastore(PraisonAISessionDataStore())
• Wired into claw/default_app.py with same integration
• Try/catch blocks ensure graceful degradation when dependencies missing

✅ Validation

• All tests pass (create, add, get, clear, delete, multiple messages)
• Import validation working correctly
• Clean error messages for missing dependencies

Next Steps (Future PRs):

• Phase 1: Replace remaining Chainlit launch points (realtime, bot, agents)
• Phase 2: Update pyproject.toml dependencies and remove Chainlit code

This unblocks the migration path while maintaining backward compatibility!

---