I’ve reviewed the Qermit-based mitigation module and identified several key issues and clean-ups

[AiPals]

Here are the proposed corrections—each illustrated with the minimal changed snippet—so you can integrate them directly.

Fixes by Jesus Carrasco

1. Fixing the “TK1” membership test in label_gates

Problem:

if labelled_command["op"]["type"] in ("TK1"):

This checks membership in the string "TK1" (i.e. {"T","K","1"}) rather than the OpType.

Correction:

from pytket.circuit import OpType

# …

if labelled_command["op"]["type"] == OpType.TK1:
    labelled_command["opgroup"] = f"Computing {comp_count}"
    comp_count += 1
elif labelled_command["op"]["type"] == OpType.CX:
    labelled_command["opgroup"] = f"Frame {frame_count}"
    frame_count += 1
else:
    raise RuntimeError(
        'Expected only TK1 or CX gates—please rebase first with PECRebase.'
    )

---

2. Consistent “noop” creation for identity Pauli

Problem:
In str_to_pauli_op and in wrapping frames, raw dicts and "noop" strings are used.

Correction:

def str_to_pauli_op(pauli_str: str) -> Op:
    assert pauli_str in ["I", "X", "Y", "Z"]
    mapping = {
        "I": OpType.Noop,
        "X": OpType.X,
        "Y": OpType.Y,
        "Z": OpType.Z,
    }
    return Op.create(mapping[pauli_str])

And in wrap_frame_gates, replace the manual dict with:

identity_op = Op.create(OpType.Noop)
# …
labelled_EM_command = {
    "op": identity_op.to_dict()["op"],
    "args": [arg],
    "opgroup": f"pre Pauli {frame_number} {qubit_i}"
}

---

3. Mapping qubits correctly (use Qubit, not Node)

Problem:

n_q_map[cc_qns[i]] = Node("q", i)

tket’s place_with_map expects Qubit-to-Qubit maps, not architecture Nodes.

Correction:

from pytket.unit_id import Qubit

# …

n_q_map = { old_q: Qubit(i) for i, old_q in enumerate(rand_cliff_circ.qubits) }
# Then use place_with_map(rand_cliff_circ, n_q_map)

---

4. Avoid reassigning circ.add_circuit (it returns None)

Problem:

circ = circ.add_circuit(cliff_circ, [...], [])

add_circuit mutates in place and returns None.

Correction:

circ.add_circuit(cliff_circ, list(range(n_qubits)), [])

---

5. Ensure full Pauli mapping in clifford_canonical_F

Problem: The doc said pauli_layer in {0,1,2,3}, but the code only handled 0,1,2.

Correction:

for i, gate in enumerate(pauli_layer):
    if gate == 1:        circ.X(i, opgroup="Clifford 1")
    elif gate == 2:      circ.Y(i, opgroup="Clifford 1")
    elif gate == 3:      circ.Z(i, opgroup="Clifford 1")
    # gate == 0: identity—no action

---

Next Steps

1. Apply these patches into your module where indicated.
2. Run your unit tests (especially around label_gates, wrapping logic, and your QubitPauliString expectations).
3. Let me know if you encounter any lingering errors or need further refinements—happy to dive deeper.