03 Jun 07:23

kjoshi07

76d5c2e

AgentForge v0.2.4 — Live-fire MCP Latest

Latest

AgentForge v0.2.4 — Live-fire MCP

The first live, Bedrock-backed MCP agent integration surfaced a
cluster of framework defects — from "the documented config does
nothing" to "every MCP tool name is rejected by the provider."
v0.2.4 fixes the whole cluster (eight bugs) and lands the one
enhancement that unblocks hosted MCP servers.

No breaking changes. Same 34 workspace packages as v0.2.3,
bumped to 0.2.4 with cross-package pins refreshed.

Highlights

MCP modules.protocols is actually wired now. Declaring an
MCP server in agentforge.yaml previously validated but did
nothing — no subprocess, no tools. The runtime now resolves
modules.protocols, starts each handler, and merges its tools
into the agent (bug-020 + bug-014). This path also fixed a
latent gap where native agent.tools from config were never
wired.
MCP tools work on every provider. Tool names are now joined
with __ (e.g. filesystem__read_file) and validated against
the ^[a-zA-Z0-9_-]{1,64}$ charset Bedrock / OpenAI / Anthropic
all enforce, with a local, actionable error before the request
leaves the process (bug-012 + bug-017). Dotted names used to
fail remotely on the first LLM call.
pip install "agentforge-py[mcp]" delivers a working MCP
runtime. The meta-package extras now chain each sister
package's vendor-SDK extra, so the SDK actually installs
(bug-015). An audit fixed 12 broken extra chains + [all].
Hosted MCP servers over HTTP (enh-001).
MCPServer.from_http(...) + serve() now serve the MCP
protocol over streamable-HTTP under uvicorn, not just stdio:
```
server = MCPServer.from_http(tools=tools, host="0.0.0.0", port=8765)
await server.serve()   # serves over HTTP instead of raising
```
Chat sessions create cleanly + terse config syntax works.
POST /sessions no longer 500s on a fresh SQLite/Postgres/Redis
store (bug-018), and the documented - faithfulness /
- presidio: {…} shorthand for evaluators and guardrails is
accepted (bug-019).

What's new

Added

MCP runtime wiring — ProtocolBridge contract +
Agent(protocol_bridges=…) + build_protocols_from_config
(feat-013). MCPBridge.attach_local_tools / MCPServer.set_tools.
agentforge_core.contracts.tool.validate_tool_name +
ToolNameInvalidError — portable tool-name validation invoked
by every provider at request-build time (bug-017).
ChatHistoryStore.create_session(...) — concrete, additive
ABC method to register a session before its first turn (bug-018).
MCP server-side HTTP transport (enh-001) — MCPServer.from_http
- serve() over streamable-HTTP (StreamableHTTPSessionManager
  under uvicorn). starlette / uvicorn arrive via
  agentforge-mcp[mcp]. SSE server transport remains deferred.

Changed

34 workspace packages bumped to 0.2.4; cross-package pins
refreshed ~= 0.2.3 → ~= 0.2.4.
MCP server from_stdio / from_http register tools at
construction; register_tools() is idempotent (bug-013).
The MCP client HTTP transport moved off the SDK's deprecated
streamablehttp_client to streamable_http_client.
agentforge-chat hard-deps aiosqlite (it was an optional
extra despite an unconditional import).

Fixed

bug-020 / bug-014 — modules.protocols.mcp never instantiated;
MCPBridge.from_config raised inside a running event loop.
bug-012 / bug-017 — MCP tool names used an illegal .
separator; provider tool-name charset undocumented + unchecked.
bug-015 — agentforge-py extras didn't pull their vendor
SDKs (12 chains + [all]; bedrock phantom extra; chat aiosqlite).
bug-019 — terse string / single-key-mapping config sugar for
evaluators + guardrails raised ValidationError.
bug-018 — POST /sessions 500'd on a fresh SQL/Redis chat
store.
bug-013 — MCPServer factories served an empty ListTools.
bug-008 — scaffolds recorded _template_version: 0.0.0+unknown
(version looked up by import name, not distribution name).

Full per-bug detail in CHANGELOG.md and the
docs/bugs/bug-0NN-*.md reports.

Breaking changes

None.

Upgrade guide

Fresh install

pip install "agentforge-py[bedrock]==0.2.4"
# MCP consumers:
pip install "agentforge-py[mcp]==0.2.4"   # now actually pulls the mcp SDK

Existing scaffolded agent

cd my-agent
agentforge upgrade --to 0.2.4
uv sync

No config changes are required. Agents already using
modules.protocols.mcp start working without edits (the config was
previously a no-op). Dotted MCP/tool names that happened to work on
Anthropic-direct should be renamed to the [a-zA-Z0-9_-] charset —
the framework now flags them locally with a suggested rewrite.

Coordinated release train

Per ADR-0015, every release bumps every in-scope workspace package
to the same version. v0.2.4 ships the same 34 packages as v0.2.3 at
0.2.4 with cross-package pins refreshed. The MCP / chat / core /
provider packages carry the substantive changes above; the rest are
version-bump-only.

Cross-language status

Python: released as v0.2.4 on PyPI.
TypeScript: not yet started (Python-first per ADR-0002;
TypeScript targets parity by 0.4).

Acknowledgements

Thanks to kjoshi — designer, implementer, reviewer, and the
downstream integrator whose live Bedrock + MCP agent surfaced this
entire cluster. Generated with Claude
Code (Anthropic) as the primary AI
co-author per the Co-Authored-By: commit trailers.

Full changelog

CHANGELOG.md — curated notes.
git log v0.2.3..v0.2.4 --oneline — every commit.
Compare view:
Scaffoldic/agentforge-py/compare/v0.2.3...v0.2.4.
PyPI:
pypi.org/project/agentforge-py/0.2.4.

What's next

v0.3 backlog unchanged — see docs/roadmap.md.
Deferred from this train: SSE MCP server transport (enh-001
phase 2, pending the upstream SDK's SSE server adapter) and
server-side expose runtime-wiring (auto-serving the agent's own
tools as an MCP server from inside the agent runtime).

Assets 2

21 May 13:42

kjoshi07

v0.2.3

0437e57

AgentForge v0.2.3 — Upgrade-flow fix (bug-007)

AgentForge v0.2.3 — Upgrade-flow fix

v0.2.2 fixed the scaffold path so fresh agentforge new agents
work end-to-end. But existing v0.2.x agents had no way to pull
those fixes — agentforge upgrade was non-functional. v0.2.3
fixes the upgrade path itself.

No new features. No breaking changes. Same 34 workspace
packages, same shipped surface as v0.2.2.

Highlights

agentforge upgrade actually works now. Existing agents
scaffolded under v0.2.1 / v0.2.2 can pull the latest template
fixes into their managed files via a single command:
```
cd my-agent
agentforge upgrade --to 0.2.3
```
Managed files refresh; forked files are preserved; any new
managed files the upgraded template introduced get added.
agentforge new now persists .agentforge-state/answers.yml.
Previously empty / missing on every fresh scaffold — which is
why upgrade was unreachable in the first place.
End-to-end validated against agents/code-reviewer/. An
agent scaffolded under buggy v0.2.1 templates was upgraded to
v0.2.3 cleanly. All six v0.2.2 scaffold fixes propagated:
agentforge-anthropic[anthropic] + python-dotenv in
pyproject.toml, strategy: "react" in agentforge.yaml,
load_dotenv() in main.py, the [project.scripts] entry,
and the updated README invocation.

What's new

Fixed — bug-007 (both halves)

Part A — agentforge new doesn't persist scaffold answers.
Copier's _answers_file directive is supposed to write
.agentforge-state/answers.yml after rendering, but doesn't
reliably for in-package templates. agentforge new now writes
the file itself with _template_name + the four resolved
template variables (project_name, project_slug,
llm_provider, description).
Part B — agentforge upgrade failed with "Template not
found". The previous implementation called Copier's
run_update, which expects the template source to be a
VCS-versioned git repo. AgentForge's templates live inside the
framework package (v0.2 compromise; the spec-aligned
agentforge-templates separate repo is a v0.4 target). v0.2.3
replaces run_update with a custom in-package upgrade that
renders to a temp directory via run_copy and copies each
non-forked managed file in place. Forked entries (set via
agentforge fork) are preserved; new template files are
added; the shared scaffold (_shared/ — runbooks +
AGENTS.md / CLAUDE.md / .cursorrules / Copilot
instructions) is re-injected so it tracks the new framework
version.

Changed

34 workspace packages bumped to 0.2.3. Cross-package
pins refreshed from ~= 0.2.2 to ~= 0.2.3.

Added

docs/bugs/bug-007-upgrade-non-functional.md — full
reproduction + root cause analysis + fix description.
Three regression tests in
packages/agentforge/tests/unit/test_scaffold_state.py:
- test_new_writes_answers_yml
- test_upgrade_refreshes_managed_files
- test_upgrade_preserves_forked_files

Breaking changes

None.

Upgrade guide

From v0.2.2 — fresh install

pip install "agentforge-py[anthropic]==0.2.3"

From v0.2.2 — existing scaffolded agent

cd my-agent
agentforge upgrade --to 0.2.3
uv sync

For agents scaffolded under v0.2.1 or earlier that pre-date
the bug-007 fix: .agentforge-state/answers.yml is missing.
Hand-write it before upgrade:

# .agentforge-state/answers.yml
_template_name: minimal   # or whichever template the agent was scaffolded from
_template_version: 0.2.1
project_name: My Agent
project_slug: my-agent
llm_provider: anthropic   # or bedrock / openai
description: An AgentForge agent.

Then agentforge upgrade --to 0.2.3 works as documented.

Coordinated release train

Per ADR-0015, every release bumps every in-scope workspace
package to the same version. v0.2.3 ships the same 34 packages
as v0.2.2 at 0.2.3 with cross-package pins refreshed.

Acknowledgements

Thanks to kjoshi — designer, implementer, reviewer.
Generated with Claude Code
(Anthropic) as the primary AI co-author per the
Co-Authored-By: commit trailers.

Full changelog

CHANGELOG.md — curated notes.
git log v0.2.2..v0.2.3 --oneline — every commit.
Compare view:
Scaffoldic/agentforge-py/compare/v0.2.2...v0.2.3.
PyPI:
pypi.org/project/agentforge-py/0.2.3.

What's next

v0.3 backlog unchanged — see
docs/roadmap.md. v0.4 candidate: the
spec-aligned agentforge-templates separate-repo migration that
restores Copier's three-way merge for managed files (today's
v0.2.3 upgrade overwrites managed files wholesale instead of
merging — relies on the agentforge fork flag to preserve
user-edited files).

Assets 2

15 May 02:53

kjoshi07

v0.2.0

9389631

v0.2.0 — Drivers

AgentForge v0.2.0 — Drivers

The second coordinated release. v0.2 is the "complete the v0.1
surface" tag: every locked ABC from v0.1 (LLMClient,
EmbeddingClient, VectorStore, GraphStore, Reranker,
Migrator, chat history) now has at least one shipped driver in
tree, the production protocol runners for MCP and A2A are live,
vendor observability backends ship, and the deferred v0.2 follow-
ups have all landed. ADR-0015 coordinated release train: every
workspace member bumps to 0.2.0.

Highlights

5 first-party LLM provider sister packages. Pick your
provider with a single string-id swap — no caller code change.
```
from agentforge import Agent

async with Agent(model="anthropic:claude-sonnet-4-7") as agent:
    print((await agent.run("Hi")).output)
# or "openai:gpt-4o", "litellm:groq/llama-3.3-70b",
# "ollama:llama3.1", "bedrock:us.anthropic.claude-sonnet-4-5-…"
```
agentforge-anthropic declares {tools, json_mode, caching, thinking, streaming}; agentforge-openai declares {tools, json_mode, streaming, vision} plus text-embedding-3-*
Matryoshka embeddings; agentforge-voyage and agentforge- ollama ship embedding clients; agentforge-litellm routes a
single agent to 100+ underlying providers.
GitHub Copilot joins the AI-assistant scaffold. Every
agentforge new now ships .github/copilot-instructions.md
alongside AGENTS.md / CLAUDE.md / .cursorrules — so
Copilot users get the same framework-aware idioms as Claude
Code / Cursor / Aider users.
GraphRAG retrieval. Retriever(graph_expansion=…) composes
with vector / hybrid / reranker (feat-023). Combine a Neo4j
knowledge graph with a vector store and the retriever expands
query hits along graph edges before ranking.
Native hybrid (BM25 + vector) on every shipped vector
store. Postgres tsvector / SQLite FTS5 / Neo4j fulltext /
SurrealDB BM25 analyser — every driver passes
run_hybrid_search_conformance (feat-022 + feat-025). RRF
fusion, configurable from agentforge.yaml > retrieval:.
Reranker contract + 4 vendor sister packages.
agentforge-reranker-sentence-transformers (local CrossEncoder),
agentforge-reranker-cohere, agentforge-reranker-voyage,
agentforge-reranker-mixedbread (feat-021). Wire from YAML
under retrieval.reranker: and rerank top-N candidates.
Sentence-window streaming guardrails. safety_mode: "sentence-window" buffers per-token streams at sentence
boundaries, runs OutputValidator.check_output on each
sentence, and only emits validated chunks downstream.
Tradeoff: small latency hit at sentence boundaries; gain:
PII or unsafe content never reaches the client unredacted
(feat-020 v0.3 polish).
Schema migrations framework. Forward up migrations across
Postgres / SQLite / Neo4j / SurrealDB drivers + parameterised
migrations for dimension-sensitive vector schemas (feat-024).
agentforge db migrate walks the migrations directory.
Production protocol runners. MCP stdio + HTTP/SSE runner
(feat-013 v0.2) and A2A HTTP + per-token streaming +
discovery (feat-014 v0.2 + v0.3) — both swap in for the v0.1
fake runners with no caller code change.
Vendor observability backends. agentforge-langfuse,
agentforge-phoenix, agentforge-evidently, agentforge- statsd — pick the dashboard you already operate; the
framework's on_step / on_finish hook fan-out plus v0.3
polish (child spans, A2A trace propagation, content-based PII
redaction) feeds it (feat-009 v0.2 + v0.3).
34 workspace members, every one at 0.2.0. ADR-0015
coordinated release train; every __version__ constant in
tree matches.

What's new

Added

feat-003 v0.2 — first-party LLM provider sister packages.
Five new Tier-3 modules following the Runner-Protocol +
lazy-SDK-import pattern:
- agentforge-anthropic — Anthropic native Messages API.
  Declares {tools, json_mode, caching, thinking, streaming}.
  Caching via cache_control: ephemeral breakpoints,
  extended thinking via thinking={"type": "enabled", "budget_tokens": ...}, per-token streaming via
  messages.stream().
- agentforge-openai — OpenAI chat.completions +
  text-embedding-3-* embeddings. Declares
  {tools, json_mode, streaming} (+vision for gpt-4o*).
  Matryoshka dimension override on embeddings.
- agentforge-voyage — Voyage AI embeddings
  (voyage-3-*, voyage-code-3, voyage-multimodal-3).
  Declares {matryoshka, multimodal} where applicable.
- agentforge-litellm — wraps litellm.acompletion so a
  single agent routes to 100+ underlying providers.
  Conservatively declares only {tools}.
- agentforge-ollama — local Ollama daemon via
  ollama.AsyncClient. Declares {tools, streaming}.
  OllamaEmbeddingClient requires explicit dimensions=
  since Ollama's API doesn't expose model→dim mapping.
feat-019 v0.2 — GitHub Copilot support. Every scaffolded
agent now ships .github/copilot-instructions.md as a
one-line pointer to the canonical AGENTS.md. Joins the
existing CLAUDE.md and .cursorrules pointer files;
inject_shared_scaffold picks the new file up automatically
via _shared/.github/copilot-instructions.md. Unit test in
test_new_cmd.py::test_scaffold_ships_ai_assistant_instructions
guards against future drift.
feat-019 v0.2 polish — 5 new runbooks auto-discovered by
agentforge docs: 17-add-reranker.md,
18-add-hybrid-search.md, 19-add-graphrag.md,
20-apply-schema-migrations.md,
21-use-streaming-guardrails.md.
feat-021 — Reranker abstraction + 4 vendor sister packages.
Reranker ABC + RerankResult + RetrievalConfig.reranker:
config block + build_retriever_from_config wiring.
Drivers: agentforge-reranker-sentence-transformers (local),
agentforge-reranker-cohere, agentforge-reranker-voyage,
agentforge-reranker-mixedbread.
feat-022 — BM25 + vector hybrid search.
VectorStore.lexical_search ABC + BM25 reference impl +
Retriever(mode="hybrid") with RRF fusion + agentforge.yaml
retrieval.mode: hybrid + retrieval.rrf_k: knob.
feat-022 v0.2 follow-up — native lexical paths. Postgres
uses to_tsvector + ts_rank_cd + GIN index. SQLite uses
FTS5 + BM25 ranking. Both drivers pass
run_hybrid_search_conformance against the in-memory
baseline.
feat-023 — GraphRAG hybrid retrieval. GraphExpansion
value + Retriever(graph_expansion=…) kwarg +
RetrievalConfig.graph_expansion: config block. Composes
cleanly with mode="hybrid" and reranker=….
feat-024 — Schema migrations framework. Migrator ABC +
agentforge db migrate CLI command + per-driver
implementations (Postgres / SQLite / Neo4j / SurrealDB). Down
migrations deferred to v0.3.
feat-024 v0.3 polish — parameterised migrations.
Migration files accept template parameters (vector
dimensions, table names) so the same migration drives
embeddings of different sizes without divergence.
feat-025 — Neo4jVectorStore + SurrealDB native lexical.
Neo4jVectorStore ships vector search over Neo4j native
vector indexes; SurrealDB's lexical_search switches from
the in-memory BM25 fallback to the native BM25 analyser.
feat-013 v0.2 — production MCP runner. Live MCPStdioRunner
- MCPHTTPRunner over the upstream MCP SDK. Replaces the v0.1
  FakeMCPRunner for production use; tests still inject the
  fake.
feat-014 v0.2 + v0.3 — production A2A runner + discovery +
streaming. A2AHTTPRunner over httpx + agent registry
- per-token streaming + run_id / budget header propagation.
feat-009 v0.2 — vendor observability sister packages.
- agentforge-langfuse — LangfuseHook over Langfuse SDK.
- agentforge-phoenix — Arize Phoenix OTel collector wiring.
- agentforge-evidently — drift / quality reports via
  Evidently (local JSON; Cloud dashboards deferred to v0.3).
- agentforge-statsd — counter / timer fan-out for StatsD /
  DogStatsD daemons.
feat-009 v0.3 — OTel polish. Child spans for strategy
iterations (ReAct / Plan-Execute / ToT / MultiAgent), A2A
trace-context propagation, and content-based PII redaction
in span attributes.
feat-020 v0.2 follow-up — chat history backends + Slack.
agentforge-chat-history-postgres (asyncpg), agentforge- chat-history-redis (hiredis + cross-process session lock),
agentforge-chat-slack adapter (SlackChatAdapter).
Provider-aware tokeniser landed for token-budget truncation.
feat-020 v0.3 polish — sentence-window streaming
guardrails. ChatSessionConfig.safety_mode Literal expanded
with "sentence-window"; _stream_per_token dispatches on
the mode and runs OutputValidator.check_output on each
completed sentence before emitting downstream.
"stream-then-redact" accepted by the schema as an alias
for now; true regex-inline stream-then-redact deferred to
v0.3+.
feat-002 + feat-009 v0.3 polish — strategy streams +
iteration spans. PlanExecute.stream() / ToT.stream() /
MultiAgent.stream() overrides land per-phase / per-depth
/ per-agent events. Matching strategy.iteration OTel spans.

Changed

All workspace members bumped to 0.2.0 per ADR-0015.
Every __version__ constant in tree updated to match.
agentforge.providers resolver category picks up five
new registrations at import time: anthropic, openai,
ollama, litellm (plus existing bedrock).
agentforge.embeddings picks up openai, voyage,
ollama (plus existing bedrock).
CI workflow split per-OS. .github/workflows/ci.yml
became ci-linux.yml (per-PR gate), ci-windows.yml +
ci-mac.yml (workflow_dispatch only). Cuts ~⅔ of per-PR
CI minutes; trigger Windows / macOS runs manually before
cutting a release.
**`ChatSession...

Assets 2

12 May 10:20

kjoshi07

v0.1.0

94cc355

AgentForge v0.1.0 — Foundation

The first tagged release of AgentForge. v0.1.0 ships the entire
locked-contract surface, every reasoning loop, persistence,
findings, evaluators, observability, guardrails, protocols
(MCP + A2A), pipelines, chat, scaffolding, testing, and the
operational CLI — across 18 coordinated workspace packages.

Highlights

Three-line agent from a fresh scaffold. agentforge new smoke --template minimal && cd smoke && agentforge run "Hi"
produces output against a real LLM. Six starter templates
(minimal, code-reviewer, patch-bot, docs-qa, triage,
research) ship in-wheel.
Locked contracts. Agent, Tool, MemoryStore,
VectorStore, GraphStore, LLMClient, EmbeddingClient,
Evaluator, Finding, FindingRenderer, ReasoningStrategy,
Task, ChatHistoryStore, HistoryTruncationStrategy,
AuthPolicy, InputValidator / OutputValidator /
ToolCallGate — every framework ABC is frozen behind ADR-0007.
Modules implement them; the runtime never reaches around.
Four reasoning loops, all stable from v0.1. ReActLoop
(default), PlanExecuteLoop, TreeOfThoughts,
MultiAgentSupervisor — pick by name in agentforge.yaml.
Full persistence stack. SQLite + Postgres + Neo4j +
SurrealDB drivers behind MemoryStore + VectorStore +
GraphStore, plus an InMemoryStore default and RAG
primitives.
Operational CLI. agentforge run / eval / debug /
db {migrate,backup,restore,purge,query} / health /
config {validate,show,schema} / list modules / add /
remove / swap module / new / upgrade / fork /
unfork / status / docs. Exit codes locked
(0/1/2/3/4/5).

from agentforge import Agent

async with Agent(model="bedrock:us.anthropic.claude-sonnet-4-5-20250929") as agent:
    result = await agent.run("Say hello in three words.")
    print(result.output)

What's new

Added

Agent orchestrator (agentforge-core + agentforge)
with locked constructor surface (feat-001). async with
lifecycle; record_runs=, input_validators=,
output_validators=, tool_gates=, guardrail_policy=,
pipeline= kwargs.
Reasoning strategies (feat-002): ReActLoop,
PlanExecuteLoop, TreeOfThoughts, MultiAgentSupervisor —
all in agentforge.strategies.
LLM + embedding providers (feat-003): LLMClient +
EmbeddingClient ABCs + agentforge-bedrock shipped.
Tools system (feat-004): @tool decorator + four
default tools + FakeTool.
Persistence (feat-005): MemoryStore ABC + sqlite,
postgres, neo4j, surrealdb drivers + VectorStore +
GraphStore + RAG primitives.
Evaluators (feat-006): Coverage /
FormatCompliance / RegressionVsBaseline / Consistency
deterministic graders + agentforge-eval-geval with
Correctness / Faithfulness / Groundedness /
Hallucination / Relevance / Helpfulness LLM judges.
Production rails (feat-007): BudgetPolicy,
RunContext + run_id / parent_run_id,
idempotency_key_for, RunIdFilter, FallbackChain.
Findings + renderers (feat-008): SimpleFinding,
PatchFinding, NarrativeFinding, MultiSpanFinding +
RendererRegistry + scorecard / patch-applier / markdown /
span-table renderers.
Observability (feat-009): on_step / on_finish hooks
- error isolation + JSON log format + agentforge-otel
  with OpenTelemetryHook + framework root span.
Module discovery + full module CLI (feat-010):
entry-point auto-load, Resolver.list_installed,
agentforge list / add / remove / swap module.
Scaffolding (feat-011): agentforge new with six
starter templates rendered via Copier; upgrade with
three-way merge; fork / unfork / status.
Configuration system (feat-012): agentforge.yaml
schema, env var interpolation, layered env files,
dotted-path overrides, AGENTFORGE_CONFIG /
AGENTFORGE_LOG_LEVEL, module-side schema integration,
agentforge config CLI.
MCP integration (feat-013): agentforge-mcp package —
MCPServerClient (stdio + HTTP/SSE), MCPServer exposer,
MCPBridge.from_config.
A2A protocol (feat-014): agentforge-a2a package —
agent_call(target, payload) client + A2AServer +
A2ABridge.from_config; bearer + mTLS auth; canonical
AuthPolicy in agentforge-core.
Pipeline + Task ABC (feat-015): agentforge.pipeline.Pipeline
engine with DAG validation, asyncio.Semaphore
parallelism, per-task timeouts, continue/fail modes;
PipelineFindingsTool; Agent(pipeline=) wiring.
Testing framework (feat-016): agentforge.testing
namespace (MockLLMClient, FakeTool, agent_factory,
pytest fixtures, record_llm); agentforge-testing package
(GoldenSetRunner, assert_snapshot, analyze_recording).
CLI runtime (feat-017): agentforge run (+ --replay
/ --record), eval (JSONL + JUnit), debug (REPL),
db {migrate,backup,restore,purge,query}, health.
MemoryStore.delete on the ABC; __step / __eval /
__run / __pipeline reserved categories.
Safety guardrails (feat-018): InputValidator /
OutputValidator / ToolCallGate ABCs + GuardrailEngine
- four built-in basics + four vendor sister packages (LLM
  Guard, Presidio, NeMo Guardrails, Llama Guard).
Developer experience (feat-019): three-section
managed/custom file format; inject_shared_scaffold post-
render hook copies _shared/ into every new scaffold; 16
runbooks + AGENTS.md + CLAUDE.md + .cursorrules
shipped day-one; agentforge docs CLI.
Chat agents (feat-020 v0.1 scope): agentforge-chat
package — ChatSession + InMemoryChatHistory +
SqliteChatHistory + four truncation strategies
(sliding-window / token-budget / summarise-oldest /
hybrid); agentforge-chat-http package — FastAPI REST +
WebSocket + SSE + bearer auth + cross-owner 403 + rate
limiting.

Changed

Coordinated release train per ADR-0015. Every workspace
package bumps to 0.1.0 in lockstep.
Release-notes scaffolding (PR #29) — every future tag
uses .claude/templates/release-notes.md walked through
.claude/checklists/pre-release.md.
Status field reconciliation (PR #28) — every shipped
canonical spec carries an honest Status line.

Deprecated

agentforge_chat_http.BearerAuthPolicy — kept as an
alias for the canonical
agentforge_core.contracts.auth.AuthPolicy. New code should
import the canonical name directly. Alias removal slated for
v0.3.

Removed

Nothing.

Fixed

Nothing release-blocking; ongoing fix-while-feature work is
captured per-PR.

Security

Bandit clean across every package; bandit -c pyproject.toml runs on every PR.
EnvBearerAuth(token_env_var) is a v0.1 placeholder
shipped in agentforge and agentforge-chat-http.
Production deployments are expected to implement
AuthPolicy against a real identity provider.

Breaking changes

None. v0.1.0 is the first tag; there is no prior public
surface to break.

Migration guide

Not applicable for the first release.

Coordinated release train

Per ADR-0015, every framework release bumps every in-scope
workspace package to the same minor version simultaneously.
The release train cut on 2026-05-12 bumps every in-scope
package to 0.1.0:

Package	Version	Surface change
`agentforge-core`	`0.1.0`	Locked contract surface — every ABC, value type, production-rails primitive, config schema, resolver, testing harness.
`agentforge`	`0.1.0`	Runtime: `Agent`, four reasoning loops, in-memory store, tools system, configuration loader, recording / replay, pipeline runtime, CLI, scaffolding + six templates + 16 runbooks.
`agentforge-bedrock`	`0.1.0`	First-party `LLMClient` against AWS Bedrock.
`agentforge-memory-sqlite`	`0.1.0`	`MemoryStore` + `VectorStore` over SQLite via `aiosqlite`.
`agentforge-memory-postgres`	`0.1.0`	`MemoryStore` + `VectorStore` over Postgres via `asyncpg`.
`agentforge-memory-neo4j`	`0.1.0`	`GraphStore` + `MemoryStore` over Neo4j.
`agentforge-memory-surrealdb`	`0.1.0`	`GraphStore` + `MemoryStore` over SurrealDB.
`agentforge-eval-geval`	`0.1.0`	LLM-judge engine + 6 named graders.
`agentforge-otel`	`0.1.0`	`OpenTelemetryHook` + framework root span.
`agentforge-testing`	`0.1.0`	`GoldenSetRunner`, `assert_snapshot`, `analyze_recording`.
`agentforge-guard-llmguard`	`0.1.0`	LLM Guard vendor wrapper.
`agentforge-guard-presidio`	`0.1.0`	Presidio vendor wrapper.
`agentforge-guard-nemo`	`0.1.0`	NeMo Guardrails vendor wrapper.
`agentforge-guard-llamaguard`	`0.1.0`	Llama Guard vendor wrapper.
`agentforge-mcp`	`0.1.0`	MCP client + server + bridge. Production runner pending live integration test (slated for v0.2).
`agentforge-chat`	`0.1.0`	`ChatSession` + in-memory + SQLite history + four truncation strategies.
`agentforge-chat-http`	`0.1.0`	FastAPI server (REST + WS + SSE + bearer auth).
`agentforge-a2a`	`0.1.0`	A2A client + server + bridge + bearer / mTLS auth. Production runner pending live integration test (slated for v0.2).

Cross-language status

Per ADR-0002, Python ships first during 0.x; TypeScript catches
up to parity by v0.4.

Python: released as 0.1.0 on PyPI.
TypeScript: not yet started. The typescript/agentforge-ts/
workspace is empty. TS parity to Python's v0.2 surface is
slated for v0.4.

Install / upgrade

# Hello-world install
pip install "agentforge[bedrock]==0.1.0"

# Or, from a fresh scaffold
pip install agentforge==0.1.0
agentforge new my-agent --template minimal
cd my-agent
agentforge run "Hi"

Shipped features

Spec	What landed in v0.1.0
[feat-001](../features/fe...

Assets 2

Releases: Scaffoldic/agentforge-py