Skip to main content

here-i-strand (HIS) agent module. Provides a custom Strands agent with DynamoDB status tracking, S3 session persistence and a concurrent tool executor with per-tool timeouts

Project description

Here I Strand

Library that provides a Strands agent for AWS Bedrock Agent Core with DynamoDB status tracking, S3 session persistence, and a concurrent tool executor with per-tool timeouts (HIS: here-i-strand).

Repository: github.com/tisaarus/here-i-strand

Installation

pip install here-i-strand
# or with uv
uv add here-i-strand

Quick start

You define how to load configuration (environment variables, your own settings module, etc.) and pass the values to HISAgent:

import os
from his import HISAgent

agent = HISAgent(
    bucket_name=os.environ["BUCKET_NAME"],
    status_dynamo_table_name=os.environ["STATUS_DYNAMO_TABLE_NAME"],
    session_id="my-session-id",
    name=os.environ.get("AGENT_NAME", "my-agent"),
    agent_id=os.environ.get("AGENT_ID"),  # Optional: unique identifier for the agent instance
    tools=[...],
)

Configuration

The library does not provide a settings layer. Pass bucket_name, status_dynamo_table_name, session_id, and optionally name, agent_id (and any other HISAgent arguments) from your own config:

  • Environment variables
  • A .env file loaded by your app (e.g. python-dotenv, pydantic-settings)
  • Any other configuration source you use

Project structure (development)

here-i-strand/
├── README.md
├── pyproject.toml
├── .env.example             # Example env vars (for your app)
│
├── his/                      # Main package
│   ├── __init__.py           # HISAgent, TimeoutConcurrentToolExecutor, event_loop_tracker, write_dynamo
│   ├── his.py                # HISAgent, TimeoutConcurrentToolExecutor, event_loop_tracker, write_dynamo
│   └── logging/
│       └── logging.py
│
└── tests/
    ├── conftest.py
    └── test_his.py

Main components

  • HISAgent: Strands agent with DynamoDB status tracking and optional S3-backed sessions. You pass status_dynamo_table_name, session_id, and optionally bucket_name, name, and agent_id so the tracker and session persistence use your table and bucket. Includes a default system prompt that enforces DynamoDB status reporting at key milestones. The write_dynamo tool is automatically added to the agent's tools.
  • TimeoutConcurrentToolExecutor: Tool executor with a per-invocation timeout (default: 300s); if a tool exceeds the limit, an error is returned and execution continues with the rest.
  • event_loop_tracker: Callback that writes event-loop milestones (init_event_loop, result, force_stop) to DynamoDB along with session and agent identifiers.
  • write_dynamo: Strands tool for writing custom event records to DynamoDB for agent tracking and observability. Automatically included in HISAgent. Each call creates a new append-only event item identified by a composite key (session_id, event_id).
  • DEFAULT_DYNAMODB_REPORTING_PROMPT_TEMPLATE: Template for the default system prompt that instructs the agent to report status updates to DynamoDB at key milestones such as START, TOOLING, MODEL_INVOCATION, PROGRESS, COMPLETION, and ERROR. The template is populated with the actual table_name, session_id, and agent_id values when the agent is created and explicitly instructs the agent to keep this reporting completely invisible to the end user.
  • HISBedrockThrottlingLogger: Hook provider that listens for ModelThrottledException events and writes BEDROCK_THROTTLING records to DynamoDB for observability, independent of the system prompt.
  • HISAgent.stream_async: Async streaming path that yields events from Strands and, on the final result event, persists AGENT_STATS and RESULT telemetry to DynamoDB.

Default behavior

When you create an HISAgent, the following happens automatically:

  1. System prompt: The DEFAULT_DYNAMODB_REPORTING_PROMPT_TEMPLATE is formatted with your status_dynamo_table_name, session_id, and agent_id and prepended to any custom system prompt you provide. It instructs the agent to report status at key milestones (START, TOOLING, MODEL_INVOCATION, PROGRESS, COMPLETION, ERROR) while keeping this reporting hidden from the user.
  2. Tools: The write_dynamo tool is automatically added to your tools list, enabling the agent to write status updates to DynamoDB.
  3. Session persistence: If you provide bucket_name, the agent uses S3SessionManager to persist session history. If bucket_name is omitted, no session manager is created.
  4. Event tracking: The event_loop_tracker callback records event-loop milestones to DynamoDB.
  5. Invocation telemetry: Both __call__ and stream_async log AGENT_STATS and RESULT events to DynamoDB when an invocation completes.

DynamoDB status events and schema

The library expects a DynamoDB table where status and telemetry events are written in an append-only fashion:

  • Partition key: session_id (String)
  • Sort key: event_id (String, UUID generated per event)

Each item also includes:

  • agent_id: Agent instance identifier
  • evt_type: Event type (e.g. START, TOOLING, MODEL_INVOCATION, PROGRESS, BEDROCK_THROTTLING, COMPLETION, ERROR, EVENT_LOOP, AGENT_STATS, RESULT)
  • evt_message: Human-readable message describing the event or a JSON-encoded payload
  • evt_datetime: ISO datetime of when the event was recorded

Typical high-level event types written by the agent are:

  • START: Beginning of request processing.
  • TOOLING: Before executing any external tool (includes tool name and purpose).
  • MODEL_INVOCATION: Before invoking another model or sub-agent.
  • PROGRESS: Significant progress or intermediate result checkpoints.
  • BEDROCK_THROTTLING: Bedrock throttling errors (rate/concurrency limits), including retry context.
  • AGENT_STATS: Aggregated usage metrics for the invocation (tokens, cache usage, user).
  • RESULT: Final structured result of the agent run, stored as JSON when possible (falls back to a {"raw": "<text>"} wrapper when parsing fails).
  • COMPLETION: Successful end of processing.
  • ERROR: Unexpected errors with details for debugging.

Tests

uv sync --all-groups
uv run pytest tests/ -v

Authors

License

MIT. See LICENSE.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

here_i_strand-0.2.11.tar.gz (15.7 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

here_i_strand-0.2.11-py3-none-any.whl (12.9 kB view details)

Uploaded Python 3

File details

Details for the file here_i_strand-0.2.11.tar.gz.

File metadata

  • Download URL: here_i_strand-0.2.11.tar.gz
  • Upload date:
  • Size: 15.7 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: uv/0.10.4 {"installer":{"name":"uv","version":"0.10.4","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"macOS","version":null,"id":null,"libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}

File hashes

Hashes for here_i_strand-0.2.11.tar.gz
Algorithm Hash digest
SHA256 e5bc1b5a47e16534e8b0b02612d0a82ab4195ed758dc41f04e52897c736d36f0
MD5 f48aa38202f44c0c2a453f14820e018c
BLAKE2b-256 9422b0af8c199952005cccfc52590755bcfdcbb89c44052d291f1fbd48f92563

See more details on using hashes here.

File details

Details for the file here_i_strand-0.2.11-py3-none-any.whl.

File metadata

  • Download URL: here_i_strand-0.2.11-py3-none-any.whl
  • Upload date:
  • Size: 12.9 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: uv/0.10.4 {"installer":{"name":"uv","version":"0.10.4","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"macOS","version":null,"id":null,"libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}

File hashes

Hashes for here_i_strand-0.2.11-py3-none-any.whl
Algorithm Hash digest
SHA256 b9843ff1530c9a63e6b0780aa11e0de6edd62a1bb100719e97bdf67ae22fbbe2
MD5 dcdacbd06c319daa56aa9f85b1b27ab5
BLAKE2b-256 4357cfe744c76e09cd0364c3c8ec6d68b05d313f9e36d6f312c8ecbedbbb995e

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page