Administration Guide

User Management

Location: Admin → Users

Creating Users

Click "Add User" and fill in username (unique), email, full name, role, initial password, and collection assignments.

Roles

User Lifecycle

• Deactivate — Immediately revokes all sessions. Account preserved for audit history.
• Force Logout — Revokes all active sessions without deactivating the account.
• Collection Assignment — Add or remove collection access. Changes take effect on next API request.

Collection Management

Location: Admin → Collections

Each collection is an isolated application with its own source code, documentation, search indices, tags, and user access controls.

Creating a Collection

AS/400 Connection & Extraction

Afunana connects to IBM i via JDBC to extract source code, call relationships, file structures, and field definitions. The extraction is non-invasive — no changes are made to production objects.

Connection Setup

Location: Admin → Configuration → AS/400 Connection

Prerequisites: IBM i accessible via TCP/IP, ports 8471 (JDBC) and 21 (FTP) open, a read-only user account, and the TTDOCPGM1 extraction program installed.

Click "Test Connection" to verify JDBC connectivity before extracting.

Extraction Process

Admin triggers extract → SBMJOB on IBM i → TTDOCPGM1 runs → IFS output → FTP download → Data/{collection}/

1. Afunana submits a batch job (SBMJOB) that runs TTDOCPGM1 on the IBM i
2. TTDOCPGM1 extracts source members, relationships, and metadata into CSV files on the IFS
3. Afunana polls the IFS every 60 seconds for a TTDOC_DONE flag file
4. Files are downloaded via FTP in binary mode and stored in Data/{collection}/

Extraction Output

Source Member Operations

Beyond extraction, Afunana can read and write individual source members on the IBM i for the AI change plan executor:

• Read: as400_connector.read_source_member(library, srcfile, member)
• Write: as400_connector.write_source_member(library, srcfile, member, records)
• IFS copy: as400_connector.copy_from_ifs_to_member(ifs_path, library, srcfile, member)

Building Documentation

Click "Build" on the collection. The AI pipeline processes all programs in the background:

1. Parse — CSV metadata to structured data
2. Build Documents — LLM generates program documentation JSON
3. System Overview — LLM generates architecture analysis
4. Embeddings — ChromaDB + BM25 search indices
5. Auto-Tagging — AI classifies programs by business function

Monitor progress via the floating log window. Build history shows past builds with status, timing, and error counts.

Build Controls

Specification Documents

On-demand generation of formal specs at three audience levels (business, analyst, programmer). Available for individual programs (Programs → Documents tab) and entire collections (System Overview → Documents). Documents are cached and regenerated when source changes.

Tags & Classification

Location: Admin → Tags

Tags classify programs and files for filtering and organization. Create tags with a key, Hebrew/English labels, color (8 options), and entity type (program or file).

During the build pipeline (Stage 5), the AI automatically assigns tags based on business function. Tags can be manually overridden. Users see tag pills on Programs and Files pages for filtering.

Configuration

Location: Admin → Configuration. Organized by category:

• API Keys — OpenAI, Anthropic, HuggingFace. Keys are masked; click to reveal (action is audited).
• AS/400 Connection — Host, port, credentials, SSL, FTP port, extraction timeouts. "Test Connection" button.
• LLM Pipeline — Model selection per role. Token limits, thinking budgets, parallelism.
• Security — Session timeouts, password policy, lockout thresholds.
• Audit & SIEM — Enable/disable logging, categories, severity, SIEM forwarding config, hash chain.
• Organization — Name, branding, CORS URL, default language.

Prompt Management

Location: Admin → Prompts. System prompts control AI documentation and chat behavior. Editable per group (Build Program/File/System Overview, Chat Classifier/Planner/Tools). Changes take effect on the next build or chat query.

Monitoring & Logs