epam/timebase-agent-plugins/timebase-csharp-client

TimeBase C# Client

Mission

Teach correct use of the TimeBase .NET client API: connection lifecycle, schema and binding, cursors and loaders, and QQL execution from C#. Prefer MCP-grounded stream and schema facts before generating code.

When to use this skill

Use only when the user explicitly requests C# or .NET together with TimeBase.

Typical tasks:

• typed reads/writes,
• stream creation with introspector or explicit schema,
• ExecuteQuery with bound result POCOs,
• Deltix.Timebase.* packages and feed setup,
• OAuth2 via AccessToken and MSAL,
• debugging binding, connection, or NuGet restore failures.

How to route requests

1. Confirm the user asked for C#/.NET + TimeBase, not generic .NET help.
2. When stream names, schema, symbols, or time bounds matter, use MCP discovery first (references/mcp-assisted-discovery.md).
3. Detect the target project: preserve existing .csproj TFM and package style when present.
4. Before generating connection, stream, or cursor/loader code, verify the exact API shape from the target project's package metadata, local package docs or a compile probe.
5. Pick the smallest proven reference and example for the access pattern.
6. For project setup and build verification, use references/project-setup-and-nuget.md.

Local API verification rules

• Do not invent interfaces, members, or overloads from memory when project metadata, package docs, or compile checks are available.
• If the API shape is still unclear after checking package docs and bundled references, use a small compile probe instead of guessing.

Mandatory policy

• Ground stream keys and schema from MCP or user input. Do not invent field names, type names, or stream keys.
• Prefer built-in Deltix.Timebase.Api.Messages types when schema matches, do not invent custom POCOs unnecessarily.
• Use TypeLoader for custom bound types and QQL result POCOs.
• When QQL authoring is central, use the QQL generator skill first. Return here for C# execution and binding when the user supplies final QQL.
• Use placeholder-based credentials in NuGet.config and preserve any existing naming convention the project already uses. Never hardcode Nexus credentials. If restore fails with 401, report that the configured credentials are missing or incorrect.
• Keep ITickDb and cursor lifecycle explicit (using or finally).
• Start read-only unless writes are required. Do not write to streams unless the user asked.

MCP-first workflow

When the task depends on live server context:

1. get_server_configuration
2. list_streams
3. get_stream_schema
4. get_stream_time_range and get_stream_symbols when needed
5. Generate C# using grounded names and types from those results

If MCP is unavailable, state what is missing and ask the user for stream/schema context before finalizing bound types.

Capability map

• `references/workflow-selection.md`: routing table
• `references/mcp-assisted-discovery.md`: MCP discovery order
• `references/project-setup-and-nuget.md`: packages, feed, credentials, verification
• `references/message-types-and-schema.md`: POCOs, introspector, TypeLoader
• `references/cursor-and-streams.md`: connect, select, cursors, live, reverse reads
• `references/qql-bound-queries.md`: ExecuteQuery, result POCOs
• `references/authentication.md`: MSAL + AccessToken
• `references/api-discovery.md`: API checks
• `references/debugging-and-performance.md`: NuGet, binding, auth failures