Add SciAgent scientific research flow configuration

[HussainAther]

Pull Request

Description

This PR introduces the initial SciAgent scientific research flow configuration, designed as a domain-agnostic, multi-stage research workflow. It provides the foundation for a general-purpose scientific reasoning agent within DeepCritical, complementing existing domain-specific flows such as PRIME and Bioinformatics.

The SciAgent statemachine is intended to support tasks such as:

• research question parsing
• hypothesis generation
• literature review orchestration
• evidence aggregation
• structured synthesis
• optional dataset logging (hypothesis traces, review traces, methods traces)

This PR only includes configuration and Hydra integration.
Implementation of execution nodes will follow in subsequent PRs.

---

Type of Change

• ✨ New feature (non-breaking change which adds functionality)
• 🔧 Configuration change

---

Component

• Configuration (Hydra)
• Pydantic Graph (future integration)
• Core Workflow Engine
• PRIME Flow
• Bioinformatics Flow
• DeepSearch Flow
• Tool Registry
• Agent System
• Documentation
• Tests
• Other

---

Related Issues

• Related to supporting a general-purpose multi-domain research pipeline (SciAgent)
• Enables dataset generation flows in future PRs
• Serves as prerequisite for hypothesis reasoning & systematic review tooling

---

Changes Made

• Added new configuration file:

  • configs/statemachines/flows/sciagent.yaml

• Added sciagent entry to configs/config.yaml under the flows: block

• Introduced parameters for:

  • parsing, hypothesis generation
  • literature review (via DeepSearch and RAG)
  • evidence analysis
  • synthesis (answers, critical reviews, methods)
  • optional dataset logging

• Enabled Hydra toggling of the flow via:

  uv run deepresearch flows.sciagent.enabled=true question="..."

---

Testing

• I have tested these changes locally
• All existing tests pass
• I have tested with different configurations

Test Configuration

uv sync
uv run deepresearch flows.sciagent.enabled=true \
  question="What mechanisms regulate TP53-mediated apoptosis?"

Test was successful at config-loading level:

• Hydra recognized the new flow
• No configuration errors
• Expected behavior pending implementation of execution nodes

---

Configuration Changes

• Added new configuration options
• Modified existing configuration
• Removed configuration options

Configuration Details

flows:
  sciagent:
    enabled: false
    params:
      max_iterations: 8
      trace_reasoning: true
      generate_datasets: true

Full config provided in configs/statemachines/flows/sciagent.yaml.

---

Documentation

• No documentation changes needed
• Updated README
• Updated configuration documentation
• Added code comments
• Updated examples
• Documentation to follow in next PR (once the nodes are implemented)

---

Performance Impact

• No performance impact
  (this PR only adds config files)

---

Breaking Changes

• No breaking changes

---

Checklist

• My code follows the project's style guidelines
• I have performed a self-review
• I have added configuration in the correct Hydra location
• My changes generate no new warnings
• New and existing unit tests pass locally
• PR is scoped and forward-compatible

---

Additional Notes

This PR is the first of several steps toward building the SciAgent subsystem.
Upcoming additions will include:

• SciAgentParse node
• SciAgentHypothesize node
• Literature fusion & evidence aggregation
• Dataset logging (hypothesis traces, review traces, methods)
• End-to-end flow execution support in DeepResearch/app.py

---

Reviewer Notes

Please verify:

• placement of the config under configs/statemachines/flows/
• naming conventions follow project standards
• flow toggling correctly integrates into the existing Hydra flows: block

No runtime behavior changes are introduced yet — this is configuration only.

---

Screenshots/Output

Not applicable for this PR.