feat(frontend): update frontend (#340)

* init

* have all sections on Frontend

* update models

* add gpt 5 series special param

* update

* update the full pipeline

* clean quick setup

* fix eslint errors

* clean up

* Fix frontend-backend compatibility issues

- Expand env_var_map in backend to cover all model options exposed in
  the frontend selector (GPT-5 variants, GPT-4.1 variants, o3/o4, all
  Claude 4.x variants, Gemini 2.x variants)
- Replace hardcoded http://localhost:5000 URLs with relative paths so
  the app works in any deployment environment:
  - DimensionSelectorModal.jsx: /api/suggest-dimensions
  - LogDisplay.jsx: socket.io now connects via window.location.origin

* Fix frontend logic bugs found in pipeline audit

- /api/code: backend returns HTTP 200 with success=false when code
  execution fails; frontend was only checking response.ok/data.error
  and would incorrectly advance to code_done state. Now also checks
  data.success === false and surfaces error_details to the user.

- generateChildNodes: function uses selectedNode state but was being
  called with pendingMerge.sourceNode as an argument (ignored). Added
  optional overrideNode parameter so the caller can pass a node
  directly, bypassing stale React state.

- DimensionSelector.jsx: /api/suggest-dimensions call was missing
  credentials:include and response.ok check, causing silent failures
  on backend errors.

* Set temperature=1.0 for Thinker and Reviewer to fix reasoning model errors

* fix agent sdk

Author: 4R5T

backend/app.py

@@ -9,7 +9,7 @@
 #name = "AI Scientist"
 
 # LLM Model
-model = "gpt-4o"
+model = "gpt-5-mini"
 
 # Temperature for controlling randomness in responses
 temperature = 0.75

config.template.toml

@@ -3,7 +3,6 @@
 import os
 from typing import Any, Dict
 
-import _bootstrap
 from tiny_scientist.budget_checker import BudgetChecker
 
 # Import the Coder class - assuming it's in a module called "coder"

examples/code.py

@@ -1,7 +1,6 @@
 import json
 import os
 
-import _bootstrap
 from tiny_scientist import TinyScientist
 
 scientist = TinyScientist(

examples/demo.py

@@ -3,7 +3,6 @@
 import json
 import os
 
-import _bootstrap
 from tiny_scientist.tool_impls import DrawerTool
 
 
@@ -25,8 +24,8 @@ def parse_args() -> argparse.Namespace:
     parser.add_argument(
         "--model",
         type=str,
-        default="claude-3-5-sonnet-20241022",
-        help="LLM model to use (default: claude-3-5-sonnet-20241022)",
+        default="gpt-5-mini",
+        help="LLM model to use (default: gpt-5-mini)",
     )
     parser.add_argument(
         "--output",

examples/draw.py

@@ -1,6 +1,5 @@
 import argparse
 
-import _bootstrap
 from tiny_scientist import TinyScientist
 
 

examples/hf_auto_eval.py

@@ -1,10 +1,9 @@
-import _bootstrap
 from tiny_scientist.utils.llm import create_client
 from tiny_scientist.utils.output_formatter import ACLOutputFormatter
 from tiny_scientist.utils.water_marker import WaterMarker
 
-client = create_client(model="gpt-4o")
-formatter = ACLOutputFormatter(model="gpt-4o", client=client)
+client = create_client(model="gpt-5-mini")
+formatter = ACLOutputFormatter(model="gpt-5-mini", client=client)
 watermarker = WaterMarker()
 formatter._compile_latex(
     cwd="../latex", output_pdf_path="../latex/output.pdf", timeout=1000

examples/latex_compile.py

@@ -4,7 +4,6 @@
 import os.path as osp
 from typing import Any, List
 
-import _bootstrap
 from tiny_scientist.reviewer import Reviewer
 from tiny_scientist.utils.llm import AVAILABLE_LLMS
 

examples/review.py

@@ -4,7 +4,6 @@
 import os
 from typing import Any, Dict, cast
 
-import _bootstrap
 from tiny_scientist.thinker import Thinker
 from tiny_scientist.utils.input_formatter import InputFormatter
 from tiny_scientist.utils.llm import AVAILABLE_LLMS

examples/think.py

@@ -3,7 +3,6 @@
 import json
 import os
 
-import _bootstrap
 from tiny_scientist.writer import Writer