Skip to content
Draft
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
57 commits
Select commit Hold shift + click to select a range
61bdd5d
Add LLM coding-agent REPL and goal-printing flags
strub May 26, 2026
4958ca5
[llm] add LOAD ... -trace for before/after sentence inspection
strub May 27, 2026
35cd4bf
[llm] relax +strict_bullets in REPL, add TREE and focus tag
strub May 30, 2026
c4fff8d
[llm] add FOCUS N and NEXT for explicit goal selection
strub May 30, 2026
4b973d2
[llm] add COMMIT to emit a +strict_bullets-friendly proof body
strub May 30, 2026
c575190
[llm] extract REPL into a dedicated EcLlm module
strub May 30, 2026
42f79eb
[llm] split REPL dispatcher into Parse and Dispatch submodules
strub May 30, 2026
eaccd1f
[llm] diagnose missing-argument REPL commands at parse time
strub May 31, 2026
6049408
[llm] nested TREE and dotted-path FOCUS
strub May 31, 2026
5b96c19
[llm] update CLAUDE.md with current workflow status
strub May 31, 2026
42a8be7
[llm] add -eval STR for scripted one-shot invocations
strub Jul 15, 2026
97bc206
[llm] use apply_pragma_option for loader pragmas (align with ec.ml co…
strub Jul 15, 2026
dbb11db
Add -stdlib DIR flag to override the built-in standard library
strub Jul 17, 2026
73b9ad6
[llm] scripted -eval runs exit nonzero when any command errors
strub Jul 23, 2026
075aef5
[llm] LOAD applies the easycrypt.project of the loaded file
strub Jul 23, 2026
e1c4f6c
[llm] add golden-output regression harness for the REPL
strub Aug 21, 2026
0d7f6e5
[llm] COMMIT keeps bullets when run after qed
strub Aug 21, 2026
9d7586c
[llm] COMMIT respects the LOAD prefix's open bullet frames
strub Aug 21, 2026
f086175
[llm] LOAD reports a missing file instead of an anomaly
strub Aug 21, 2026
8fd181a
[llm] queries no longer pollute the COMMIT transcript
strub Aug 21, 2026
e92c9b9
[llm] tag GOALS / TREE / COMMIT replies with the focus indicator
strub Aug 21, 2026
5f6029b
[llm] a -trace LOAD that cannot trace keeps the prefix state
strub Aug 21, 2026
a4a059b
[llm] split ecLlm into an engine core and a text front-end
strub Aug 21, 2026
2e0cfc8
[llm] core: TRY semantics via EcLlmCore.try_step
strub Aug 21, 2026
f24e723
[llm] add `easycrypt mcp`: an MCP server front-end over EcLlmCore
strub Aug 21, 2026
0212bc0
[llm] step runs every sentence of its input, not just the first
strub Aug 21, 2026
ecb8313
[llm] queries no longer spend a uuid
strub Aug 21, 2026
089c578
[llm] core computes `changed' for failures too
strub Aug 21, 2026
c1353d9
[llm] a golden harness for the MCP front-end
strub Aug 21, 2026
7871f70
[llm] parity check: two front-ends, one core
strub Aug 21, 2026
e37f645
[llm] real-client verification of `easycrypt mcp'
strub Aug 21, 2026
24a2b7f
[llm] mcp: repeat the reply text inside structuredContent
strub Aug 21, 2026
5968c0a
[llm] document the MCP front-end
strub Aug 21, 2026
40350f6
[llm] wire the llm/mcp golden harnesses into check and CI
strub Aug 21, 2026
597f625
[llm] SEARCH takes a pattern, not a command stream
strub Aug 21, 2026
cf98d3c
[llm] COMMIT emits sibling subtrees in DAG order, not typing order
strub Aug 22, 2026
565928c
[llm] escape envelope-shaped lines in REPL reply bodies
strub Aug 22, 2026
7b9bf47
[llm] `print' renders inside the reply, not around it
strub Aug 22, 2026
97b1df4
[mcp] repair non-UTF-8 bytes before they reach the wire
strub Aug 22, 2026
de213f8
[llm] LOAD rewinds the include path instead of growing it
strub Aug 22, 2026
3b401c9
[llm] ec_try restores the exact pre-call state, forward included
strub Aug 22, 2026
d9fb771
[llm] correct the documented meaning of FOCUS and NEXT
strub Aug 22, 2026
6bd9583
[llm] COMMIT renders each proof of a session on its own
strub Aug 22, 2026
d0a0831
[llm] QUIT and `exit.` no longer defeat the -eval exit status
strub Aug 22, 2026
197d50f
[llm] LOAD honours `upto` for every sentence kind, not just commands
strub Aug 22, 2026
30bf651
[llm] drop the -upto/-lastgoals machinery the batch llm mode left behind
strub Aug 22, 2026
f7f5cec
[llm] cover ec_load's nosmt and trace options on the MCP side
strub Aug 22, 2026
1172abc
[llm] `pragma restart.` drops the checkpoint table too
strub Aug 22, 2026
1011e44
[llm] document the front-end-facing half of ecCommands' interface
strub Aug 22, 2026
d111e90
[llm] count open goals for the focus tag instead of rendering them
strub Aug 22, 2026
d288ced
[llm] share the argument checks both front-ends were duplicating
strub Aug 23, 2026
1bd1cbe
[mcp] stop naming the JSON constructors yojson 3 removed
strub Sep 1, 2026
4c6b294
[llm] re-record the print goldens for the new locate prefix
strub Sep 4, 2026
76288f7
[llm] LOAD -noproof: admit the prefix's lemmas instead of proving them
bgregoir Sep 7, 2026
ed67c47
[llm] keep elaborated theories across a LOAD, so reloading is cheap
bgregoir Sep 7, 2026
283d37b
[llm] STRICT: stop the session at a failure instead of drifting past it
bgregoir Sep 7, 2026
fb70253
[llm] mcp -sessions: one engine per agent behind one MCP server
bgregoir Sep 10, 2026
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion .github/workflows/ci.yml
Original file line number Diff line number Diff line change
Expand Up @@ -162,7 +162,7 @@ jobs:
strategy:
fail-fast: false
matrix:
target: [unit, stdlib, examples]
target: [unit, stdlib, examples, test-llm, test-mcp]
steps:
- uses: actions/checkout@v4
- uses: actions/download-artifact@v4
Expand Down
21 changes: 20 additions & 1 deletion Makefile
Original file line number Diff line number Diff line change
Expand Up @@ -11,6 +11,15 @@ CHECK += --bin=./ec.native
CHECK += --jobs="$(ECJOBS)"
CHECK += $(foreach arg,$(ECARGS),--bin-args="$(arg)")
CHECK += $(ECEXTRA) config/tests.config
LLMCHECK := scripts/testing/llm-golden
LLMCHECK += --bin=./ec.native
LLMWARM := scripts/testing/llm-warm-reload
LLMWARM += --bin=./ec.native
MCPCHECK := scripts/testing/mcp-golden
MCPCHECK += --bin=./ec.native
MCPPARITY := scripts/testing/mcp-parity
MCPSESSIONS := scripts/testing/mcp-sessions
MCPPARITY += --bin=./ec.native
NIX ?= nix --extra-experimental-features "nix-command flakes"
PROFILE ?= dev

Expand All @@ -20,6 +29,7 @@ UNAME_S = $(shell uname -s)

# --------------------------------------------------------------------
.PHONY: default build byte native tests check examples
.PHONY: test-llm test-mcp
.PHONY: nix-build nix-build-with-provers nix-develop
.PHONY: clean install uninstall

Expand Down Expand Up @@ -49,7 +59,16 @@ stdlib: build
examples: build
$(CHECK) examples mee-cbc

check: unit stdlib examples
test-llm: build
$(LLMCHECK)
$(LLMWARM)

test-mcp: build
$(MCPCHECK)
$(MCPPARITY)
$(MCPSESSIONS)

check: unit stdlib examples test-llm test-mcp
@true

nix-build:
Expand Down
9 changes: 9 additions & 0 deletions README.md
Original file line number Diff line number Diff line change
Expand Up @@ -166,6 +166,15 @@ with proof scripts). At present, the only available front-end is based on Emacs'
[Proof General](https://github.com/ProofGeneral/PG).
However, a front-end for VSCode is currently in development.

Besides these, EasyCrypt ships an interface aimed at LLM agents rather
than at humans: `easycrypt llm`, an interactive REPL speaking a
machine-friendly protocol, and `easycrypt mcp`, a
[Model Context Protocol](https://modelcontextprotocol.io/) server over
stdio (`easycrypt mcp -sessions` serves one engine per named session,
for clients that run several agents at once). Both drive the same
proof engine, and both are documented in
[doc/llm/CLAUDE.md](doc/llm/CLAUDE.md).

### Proof-General (Emacs)

EasyCrypt mode has been integrated upstream. Please, go
Expand Down
Loading
Loading